]>
Commit | Line | Data |
---|---|---|
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 | 39 | static 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 | 43 | static 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 | 48 | static 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 | 53 | static 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 | 71 | static 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. */ | |
102 | enum 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 | 144 | static 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 | |
b3cad287 | 205 | * sLA -> sTW Last ACK detected (RFC5961 challenged) |
9fb9cbb1 YK |
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 | |
b3cad287 | 264 | * sLA -> sTW Last ACK detected (RFC5961 challenged) |
9fb9cbb1 YK |
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 |
274 | static inline struct nf_tcp_net *tcp_pernet(struct net *net) |
275 | { | |
276 | return &net->ct.nf_ct_proto.tcp; | |
277 | } | |
278 | ||
09f263cd | 279 | static bool tcp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff, |
a31f1adc | 280 | struct net *net, struct nf_conntrack_tuple *tuple) |
9fb9cbb1 | 281 | { |
82f568fc JE |
282 | const struct tcphdr *hp; |
283 | struct tcphdr _hdr; | |
9fb9cbb1 | 284 | |
e5e693ab GF |
285 | /* Actually only need first 4 bytes to get ports. */ |
286 | hp = skb_header_pointer(skb, dataoff, 4, &_hdr); | |
9fb9cbb1 | 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 |
296 | static 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 | ||
9fb9cbb1 | 304 | /* Print out the private part of the conntrack. */ |
37246a58 | 305 | static void tcp_print_conntrack(struct seq_file *s, struct nf_conn *ct) |
9fb9cbb1 | 306 | { |
a163f2cb | 307 | seq_printf(s, "%s ", tcp_conntrack_names[ct->proto.tcp.state]); |
9fb9cbb1 YK |
308 | } |
309 | ||
310 | static unsigned int get_conntrack_index(const struct tcphdr *tcph) | |
311 | { | |
312 | if (tcph->rst) return TCP_RST_SET; | |
313 | else if (tcph->syn) return (tcph->ack ? TCP_SYNACK_SET : TCP_SYN_SET); | |
314 | else if (tcph->fin) return TCP_FIN_SET; | |
315 | else if (tcph->ack) return TCP_ACK_SET; | |
316 | else return TCP_NONE_SET; | |
317 | } | |
318 | ||
319 | /* TCP connection tracking based on 'Real Stateful TCP Packet Filtering | |
320 | in IP Filter' by Guido van Rooij. | |
601e68e1 | 321 | |
631dd1a8 JM |
322 | http://www.sane.nl/events/sane2000/papers.html |
323 | http://www.darkart.com/mirrors/www.obfuscation.org/ipf/ | |
601e68e1 | 324 | |
9fb9cbb1 YK |
325 | The boundaries and the conditions are changed according to RFC793: |
326 | the packet must intersect the window (i.e. segments may be | |
327 | after the right or before the left edge) and thus receivers may ACK | |
328 | segments after the right edge of the window. | |
329 | ||
601e68e1 | 330 | td_maxend = max(sack + max(win,1)) seen in reply packets |
9fb9cbb1 YK |
331 | td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets |
332 | td_maxwin += seq + len - sender.td_maxend | |
333 | if seq + len > sender.td_maxend | |
334 | td_end = max(seq + len) seen in sent packets | |
601e68e1 | 335 | |
9fb9cbb1 YK |
336 | I. Upper bound for valid data: seq <= sender.td_maxend |
337 | II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin | |
84ebe1cd JK |
338 | III. Upper bound for valid (s)ack: sack <= receiver.td_end |
339 | IV. Lower bound for valid (s)ack: sack >= receiver.td_end - MAXACKWINDOW | |
9fb9cbb1 | 340 | |
84ebe1cd JK |
341 | where sack is the highest right edge of sack block found in the packet |
342 | or ack in the case of packet without SACK option. | |
9fb9cbb1 | 343 | |
84ebe1cd | 344 | The upper bound limit for a valid (s)ack is not ignored - |
601e68e1 | 345 | we doesn't have to deal with fragments. |
9fb9cbb1 YK |
346 | */ |
347 | ||
348 | static inline __u32 segment_seq_plus_len(__u32 seq, | |
349 | size_t len, | |
350 | unsigned int dataoff, | |
82f568fc | 351 | const struct tcphdr *tcph) |
9fb9cbb1 YK |
352 | { |
353 | /* XXX Should I use payload length field in IP/IPv6 header ? | |
354 | * - YK */ | |
355 | return (seq + len - dataoff - tcph->doff*4 | |
356 | + (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0)); | |
357 | } | |
601e68e1 | 358 | |
9fb9cbb1 YK |
359 | /* Fixme: what about big packets? */ |
360 | #define MAXACKWINCONST 66000 | |
361 | #define MAXACKWINDOW(sender) \ | |
362 | ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \ | |
363 | : MAXACKWINCONST) | |
601e68e1 | 364 | |
9fb9cbb1 YK |
365 | /* |
366 | * Simplified tcp_parse_options routine from tcp_input.c | |
367 | */ | |
368 | static void tcp_options(const struct sk_buff *skb, | |
369 | unsigned int dataoff, | |
82f568fc | 370 | const struct tcphdr *tcph, |
9fb9cbb1 YK |
371 | struct ip_ct_tcp_state *state) |
372 | { | |
373 | unsigned char buff[(15 * 4) - sizeof(struct tcphdr)]; | |
82f568fc | 374 | const unsigned char *ptr; |
9fb9cbb1 YK |
375 | int length = (tcph->doff*4) - sizeof(struct tcphdr); |
376 | ||
377 | if (!length) | |
378 | return; | |
379 | ||
380 | ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr), | |
381 | length, buff); | |
382 | BUG_ON(ptr == NULL); | |
383 | ||
601e68e1 | 384 | state->td_scale = |
9fb9cbb1 YK |
385 | state->flags = 0; |
386 | ||
387 | while (length > 0) { | |
388 | int opcode=*ptr++; | |
389 | int opsize; | |
390 | ||
391 | switch (opcode) { | |
392 | case TCPOPT_EOL: | |
393 | return; | |
394 | case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */ | |
395 | length--; | |
396 | continue; | |
397 | default: | |
644c7e48 JK |
398 | if (length < 2) |
399 | return; | |
9fb9cbb1 YK |
400 | opsize=*ptr++; |
401 | if (opsize < 2) /* "silly options" */ | |
402 | return; | |
403 | if (opsize > length) | |
4a5cc84a | 404 | return; /* don't parse partial options */ |
9fb9cbb1 | 405 | |
601e68e1 | 406 | if (opcode == TCPOPT_SACK_PERM |
9fb9cbb1 YK |
407 | && opsize == TCPOLEN_SACK_PERM) |
408 | state->flags |= IP_CT_TCP_FLAG_SACK_PERM; | |
409 | else if (opcode == TCPOPT_WINDOW | |
410 | && opsize == TCPOLEN_WINDOW) { | |
411 | state->td_scale = *(u_int8_t *)ptr; | |
412 | ||
122868b3 GF |
413 | if (state->td_scale > TCP_MAX_WSCALE) |
414 | state->td_scale = TCP_MAX_WSCALE; | |
415 | ||
9fb9cbb1 YK |
416 | state->flags |= |
417 | IP_CT_TCP_FLAG_WINDOW_SCALE; | |
418 | } | |
419 | ptr += opsize - 2; | |
420 | length -= opsize; | |
421 | } | |
422 | } | |
423 | } | |
424 | ||
425 | static void tcp_sack(const struct sk_buff *skb, unsigned int dataoff, | |
82f568fc | 426 | const struct tcphdr *tcph, __u32 *sack) |
9fb9cbb1 | 427 | { |
601e68e1 | 428 | unsigned char buff[(15 * 4) - sizeof(struct tcphdr)]; |
82f568fc | 429 | const unsigned char *ptr; |
9fb9cbb1 YK |
430 | int length = (tcph->doff*4) - sizeof(struct tcphdr); |
431 | __u32 tmp; | |
432 | ||
433 | if (!length) | |
434 | return; | |
435 | ||
436 | ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr), | |
437 | length, buff); | |
438 | BUG_ON(ptr == NULL); | |
439 | ||
440 | /* Fast path for timestamp-only option */ | |
bb9fc373 | 441 | if (length == TCPOLEN_TSTAMP_ALIGNED |
8f05ce91 YH |
442 | && *(__be32 *)ptr == htonl((TCPOPT_NOP << 24) |
443 | | (TCPOPT_NOP << 16) | |
444 | | (TCPOPT_TIMESTAMP << 8) | |
445 | | TCPOLEN_TIMESTAMP)) | |
9fb9cbb1 YK |
446 | return; |
447 | ||
448 | while (length > 0) { | |
449 | int opcode = *ptr++; | |
450 | int opsize, i; | |
451 | ||
452 | switch (opcode) { | |
453 | case TCPOPT_EOL: | |
454 | return; | |
455 | case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */ | |
456 | length--; | |
457 | continue; | |
458 | default: | |
644c7e48 JK |
459 | if (length < 2) |
460 | return; | |
9fb9cbb1 YK |
461 | opsize = *ptr++; |
462 | if (opsize < 2) /* "silly options" */ | |
463 | return; | |
464 | if (opsize > length) | |
4a5cc84a | 465 | return; /* don't parse partial options */ |
9fb9cbb1 | 466 | |
601e68e1 YH |
467 | if (opcode == TCPOPT_SACK |
468 | && opsize >= (TCPOLEN_SACK_BASE | |
469 | + TCPOLEN_SACK_PERBLOCK) | |
470 | && !((opsize - TCPOLEN_SACK_BASE) | |
471 | % TCPOLEN_SACK_PERBLOCK)) { | |
472 | for (i = 0; | |
473 | i < (opsize - TCPOLEN_SACK_BASE); | |
474 | i += TCPOLEN_SACK_PERBLOCK) { | |
534f81a5 | 475 | tmp = get_unaligned_be32((__be32 *)(ptr+i)+1); |
9fb9cbb1 YK |
476 | |
477 | if (after(tmp, *sack)) | |
478 | *sack = tmp; | |
479 | } | |
480 | return; | |
481 | } | |
482 | ptr += opsize - 2; | |
483 | length -= opsize; | |
484 | } | |
485 | } | |
486 | } | |
487 | ||
09f263cd JE |
488 | static bool tcp_in_window(const struct nf_conn *ct, |
489 | struct ip_ct_tcp *state, | |
490 | enum ip_conntrack_dir dir, | |
491 | unsigned int index, | |
492 | const struct sk_buff *skb, | |
493 | unsigned int dataoff, | |
494 | const struct tcphdr *tcph, | |
76108cea | 495 | u_int8_t pf) |
9fb9cbb1 | 496 | { |
c2a2c7e0 | 497 | struct net *net = nf_ct_net(ct); |
d2ba1fde | 498 | struct nf_tcp_net *tn = tcp_pernet(net); |
9fb9cbb1 YK |
499 | struct ip_ct_tcp_state *sender = &state->seen[dir]; |
500 | struct ip_ct_tcp_state *receiver = &state->seen[!dir]; | |
82f568fc | 501 | const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple; |
9fb9cbb1 | 502 | __u32 seq, ack, sack, end, win, swin; |
2d89c68a | 503 | s32 receiver_offset; |
356d7d88 | 504 | bool res, in_recv_win; |
9fb9cbb1 YK |
505 | |
506 | /* | |
507 | * Get the required data from the packet. | |
508 | */ | |
509 | seq = ntohl(tcph->seq); | |
510 | ack = sack = ntohl(tcph->ack_seq); | |
511 | win = ntohs(tcph->window); | |
512 | end = segment_seq_plus_len(seq, skb->len, dataoff, tcph); | |
513 | ||
514 | if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM) | |
515 | tcp_sack(skb, dataoff, tcph, &sack); | |
516 | ||
f9dd09c7 | 517 | /* Take into account NAT sequence number mangling */ |
41d73ec0 | 518 | receiver_offset = nf_ct_seq_offset(ct, !dir, ack - 1); |
f9dd09c7 JK |
519 | ack -= receiver_offset; |
520 | sack -= receiver_offset; | |
521 | ||
0d53778e PM |
522 | pr_debug("tcp_in_window: START\n"); |
523 | pr_debug("tcp_in_window: "); | |
3c9fba65 | 524 | nf_ct_dump_tuple(tuple); |
f9dd09c7 JK |
525 | pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n", |
526 | seq, ack, receiver_offset, sack, receiver_offset, win, end); | |
0d53778e PM |
527 | pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i " |
528 | "receiver end=%u maxend=%u maxwin=%u scale=%i\n", | |
529 | sender->td_end, sender->td_maxend, sender->td_maxwin, | |
530 | sender->td_scale, | |
531 | receiver->td_end, receiver->td_maxend, receiver->td_maxwin, | |
532 | receiver->td_scale); | |
9fb9cbb1 | 533 | |
874ab923 | 534 | if (sender->td_maxwin == 0) { |
9fb9cbb1 YK |
535 | /* |
536 | * Initialize sender data. | |
537 | */ | |
874ab923 | 538 | if (tcph->syn) { |
9fb9cbb1 | 539 | /* |
874ab923 JK |
540 | * SYN-ACK in reply to a SYN |
541 | * or SYN from reply direction in simultaneous open. | |
9fb9cbb1 | 542 | */ |
601e68e1 | 543 | sender->td_end = |
9fb9cbb1 YK |
544 | sender->td_maxend = end; |
545 | sender->td_maxwin = (win == 0 ? 1 : win); | |
546 | ||
547 | tcp_options(skb, dataoff, tcph, sender); | |
601e68e1 | 548 | /* |
9fb9cbb1 YK |
549 | * RFC 1323: |
550 | * Both sides must send the Window Scale option | |
551 | * to enable window scaling in either direction. | |
552 | */ | |
553 | if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE | |
554 | && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE)) | |
601e68e1 | 555 | sender->td_scale = |
9fb9cbb1 | 556 | receiver->td_scale = 0; |
874ab923 JK |
557 | if (!tcph->ack) |
558 | /* Simultaneous open */ | |
559 | return true; | |
9fb9cbb1 YK |
560 | } else { |
561 | /* | |
562 | * We are in the middle of a connection, | |
563 | * its history is lost for us. | |
564 | * Let's try to use the data from the packet. | |
601e68e1 | 565 | */ |
9fb9cbb1 | 566 | sender->td_end = end; |
6ee0b693 CG |
567 | swin = win << sender->td_scale; |
568 | sender->td_maxwin = (swin == 0 ? 1 : swin); | |
9fb9cbb1 | 569 | sender->td_maxend = end + sender->td_maxwin; |
fac42a9a PNA |
570 | /* |
571 | * We haven't seen traffic in the other direction yet | |
572 | * but we have to tweak window tracking to pass III | |
573 | * and IV until that happens. | |
574 | */ | |
575 | if (receiver->td_maxwin == 0) | |
576 | receiver->td_end = receiver->td_maxend = sack; | |
9fb9cbb1 YK |
577 | } |
578 | } else if (((state->state == TCP_CONNTRACK_SYN_SENT | |
579 | && dir == IP_CT_DIR_ORIGINAL) | |
580 | || (state->state == TCP_CONNTRACK_SYN_RECV | |
581 | && dir == IP_CT_DIR_REPLY)) | |
582 | && after(end, sender->td_end)) { | |
583 | /* | |
584 | * RFC 793: "if a TCP is reinitialized ... then it need | |
601e68e1 | 585 | * not wait at all; it must only be sure to use sequence |
9fb9cbb1 YK |
586 | * numbers larger than those recently used." |
587 | */ | |
588 | sender->td_end = | |
589 | sender->td_maxend = end; | |
590 | sender->td_maxwin = (win == 0 ? 1 : win); | |
591 | ||
592 | tcp_options(skb, dataoff, tcph, sender); | |
593 | } | |
594 | ||
595 | if (!(tcph->ack)) { | |
596 | /* | |
597 | * If there is no ACK, just pretend it was set and OK. | |
598 | */ | |
599 | ack = sack = receiver->td_end; | |
601e68e1 YH |
600 | } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) == |
601 | (TCP_FLAG_ACK|TCP_FLAG_RST)) | |
9fb9cbb1 YK |
602 | && (ack == 0)) { |
603 | /* | |
604 | * Broken TCP stacks, that set ACK in RST packets as well | |
605 | * with zero ack value. | |
606 | */ | |
607 | ack = sack = receiver->td_end; | |
608 | } | |
609 | ||
4a70bbfa | 610 | if (tcph->rst && seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT) |
9fb9cbb1 | 611 | /* |
4a70bbfa | 612 | * RST sent answering SYN. |
9fb9cbb1 YK |
613 | */ |
614 | seq = end = sender->td_end; | |
615 | ||
0d53778e | 616 | pr_debug("tcp_in_window: "); |
3c9fba65 | 617 | nf_ct_dump_tuple(tuple); |
f9dd09c7 JK |
618 | pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n", |
619 | seq, ack, receiver_offset, sack, receiver_offset, win, end); | |
0d53778e PM |
620 | pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i " |
621 | "receiver end=%u maxend=%u maxwin=%u scale=%i\n", | |
622 | sender->td_end, sender->td_maxend, sender->td_maxwin, | |
623 | sender->td_scale, | |
624 | receiver->td_end, receiver->td_maxend, receiver->td_maxwin, | |
625 | receiver->td_scale); | |
626 | ||
356d7d88 YC |
627 | /* Is the ending sequence in the receive window (if available)? */ |
628 | in_recv_win = !receiver->td_maxwin || | |
629 | after(end, sender->td_end - receiver->td_maxwin - 1); | |
630 | ||
0d53778e PM |
631 | pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n", |
632 | before(seq, sender->td_maxend + 1), | |
356d7d88 | 633 | (in_recv_win ? 1 : 0), |
0d53778e | 634 | before(sack, receiver->td_end + 1), |
84ebe1cd | 635 | after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)); |
9fb9cbb1 | 636 | |
a09113c2 | 637 | if (before(seq, sender->td_maxend + 1) && |
356d7d88 | 638 | in_recv_win && |
a09113c2 | 639 | before(sack, receiver->td_end + 1) && |
84ebe1cd | 640 | after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)) { |
601e68e1 | 641 | /* |
9fb9cbb1 YK |
642 | * Take into account window scaling (RFC 1323). |
643 | */ | |
644 | if (!tcph->syn) | |
645 | win <<= sender->td_scale; | |
646 | ||
647 | /* | |
648 | * Update sender data. | |
649 | */ | |
650 | swin = win + (sack - ack); | |
651 | if (sender->td_maxwin < swin) | |
652 | sender->td_maxwin = swin; | |
ae375044 | 653 | if (after(end, sender->td_end)) { |
9fb9cbb1 | 654 | sender->td_end = end; |
ae375044 PM |
655 | sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED; |
656 | } | |
bfcaa502 JK |
657 | if (tcph->ack) { |
658 | if (!(sender->flags & IP_CT_TCP_FLAG_MAXACK_SET)) { | |
659 | sender->td_maxack = ack; | |
660 | sender->flags |= IP_CT_TCP_FLAG_MAXACK_SET; | |
661 | } else if (after(ack, sender->td_maxack)) | |
662 | sender->td_maxack = ack; | |
663 | } | |
664 | ||
9fb9cbb1 YK |
665 | /* |
666 | * Update receiver data. | |
667 | */ | |
fac42a9a | 668 | if (receiver->td_maxwin != 0 && after(end, sender->td_maxend)) |
9fb9cbb1 YK |
669 | receiver->td_maxwin += end - sender->td_maxend; |
670 | if (after(sack + win, receiver->td_maxend - 1)) { | |
671 | receiver->td_maxend = sack + win; | |
672 | if (win == 0) | |
673 | receiver->td_maxend++; | |
674 | } | |
ae375044 PM |
675 | if (ack == receiver->td_end) |
676 | receiver->flags &= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED; | |
9fb9cbb1 | 677 | |
601e68e1 | 678 | /* |
9fb9cbb1 YK |
679 | * Check retransmissions. |
680 | */ | |
681 | if (index == TCP_ACK_SET) { | |
682 | if (state->last_dir == dir | |
683 | && state->last_seq == seq | |
684 | && state->last_ack == ack | |
c1fe3ca5 GH |
685 | && state->last_end == end |
686 | && state->last_win == win) | |
9fb9cbb1 YK |
687 | state->retrans++; |
688 | else { | |
689 | state->last_dir = dir; | |
690 | state->last_seq = seq; | |
691 | state->last_ack = ack; | |
692 | state->last_end = end; | |
c1fe3ca5 | 693 | state->last_win = win; |
9fb9cbb1 YK |
694 | state->retrans = 0; |
695 | } | |
696 | } | |
09f263cd | 697 | res = true; |
9fb9cbb1 | 698 | } else { |
09f263cd | 699 | res = false; |
a09113c2 | 700 | if (sender->flags & IP_CT_TCP_FLAG_BE_LIBERAL || |
d2ba1fde | 701 | tn->tcp_be_liberal) |
09f263cd | 702 | res = true; |
c2a2c7e0 | 703 | if (!res && LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 | 704 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, |
9fb9cbb1 YK |
705 | "nf_ct_tcp: %s ", |
706 | before(seq, sender->td_maxend + 1) ? | |
356d7d88 | 707 | in_recv_win ? |
9fb9cbb1 | 708 | before(sack, receiver->td_end + 1) ? |
f9dd09c7 | 709 | after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1) ? "BUG" |
9fb9cbb1 YK |
710 | : "ACK is under the lower bound (possible overly delayed ACK)" |
711 | : "ACK is over the upper bound (ACKed data not seen yet)" | |
712 | : "SEQ is under the lower bound (already ACKed data retransmitted)" | |
713 | : "SEQ is over the upper bound (over the window of the receiver)"); | |
601e68e1 YH |
714 | } |
715 | ||
09f263cd | 716 | pr_debug("tcp_in_window: res=%u sender end=%u maxend=%u maxwin=%u " |
0d53778e PM |
717 | "receiver end=%u maxend=%u maxwin=%u\n", |
718 | res, sender->td_end, sender->td_maxend, sender->td_maxwin, | |
719 | receiver->td_end, receiver->td_maxend, receiver->td_maxwin); | |
9fb9cbb1 YK |
720 | |
721 | return res; | |
722 | } | |
723 | ||
5c8ce7c9 | 724 | /* table of valid flag combinations - PUSH, ECE and CWR are always valid */ |
a3433f35 CG |
725 | static const u8 tcp_valid_flags[(TCPHDR_FIN|TCPHDR_SYN|TCPHDR_RST|TCPHDR_ACK| |
726 | TCPHDR_URG) + 1] = | |
9fb9cbb1 | 727 | { |
a3433f35 CG |
728 | [TCPHDR_SYN] = 1, |
729 | [TCPHDR_SYN|TCPHDR_URG] = 1, | |
730 | [TCPHDR_SYN|TCPHDR_ACK] = 1, | |
731 | [TCPHDR_RST] = 1, | |
732 | [TCPHDR_RST|TCPHDR_ACK] = 1, | |
733 | [TCPHDR_FIN|TCPHDR_ACK] = 1, | |
734 | [TCPHDR_FIN|TCPHDR_ACK|TCPHDR_URG] = 1, | |
735 | [TCPHDR_ACK] = 1, | |
736 | [TCPHDR_ACK|TCPHDR_URG] = 1, | |
9fb9cbb1 YK |
737 | }; |
738 | ||
739 | /* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */ | |
8fea97ec | 740 | static int tcp_error(struct net *net, struct nf_conn *tmpl, |
74c51a14 | 741 | struct sk_buff *skb, |
9fb9cbb1 | 742 | unsigned int dataoff, |
76108cea | 743 | u_int8_t pf, |
96f6bf82 | 744 | unsigned int hooknum) |
9fb9cbb1 | 745 | { |
82f568fc JE |
746 | const struct tcphdr *th; |
747 | struct tcphdr _tcph; | |
9fb9cbb1 YK |
748 | unsigned int tcplen = skb->len - dataoff; |
749 | u_int8_t tcpflags; | |
750 | ||
751 | /* Smaller that minimal TCP header? */ | |
752 | th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph); | |
753 | if (th == NULL) { | |
c2a2c7e0 | 754 | if (LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 | 755 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, |
9fb9cbb1 YK |
756 | "nf_ct_tcp: short packet "); |
757 | return -NF_ACCEPT; | |
601e68e1 YH |
758 | } |
759 | ||
9fb9cbb1 YK |
760 | /* Not whole TCP header or malformed packet */ |
761 | if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) { | |
c2a2c7e0 | 762 | if (LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 | 763 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, |
9fb9cbb1 YK |
764 | "nf_ct_tcp: truncated/malformed packet "); |
765 | return -NF_ACCEPT; | |
766 | } | |
601e68e1 | 767 | |
9fb9cbb1 YK |
768 | /* Checksum invalid? Ignore. |
769 | * We skip checking packets on the outgoing path | |
84fa7933 | 770 | * because the checksum is assumed to be correct. |
9fb9cbb1 YK |
771 | */ |
772 | /* FIXME: Source route IP option packets --RR */ | |
c04d0552 | 773 | if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING && |
96f6bf82 | 774 | nf_checksum(skb, hooknum, dataoff, IPPROTO_TCP, pf)) { |
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: bad TCP checksum "); |
778 | return -NF_ACCEPT; | |
779 | } | |
780 | ||
781 | /* Check TCP flags. */ | |
a3433f35 | 782 | tcpflags = (tcp_flag_byte(th) & ~(TCPHDR_ECE|TCPHDR_CWR|TCPHDR_PSH)); |
9fb9cbb1 | 783 | if (!tcp_valid_flags[tcpflags]) { |
c2a2c7e0 | 784 | if (LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 | 785 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, |
9fb9cbb1 YK |
786 | "nf_ct_tcp: invalid TCP flag combination "); |
787 | return -NF_ACCEPT; | |
788 | } | |
789 | ||
790 | return NF_ACCEPT; | |
791 | } | |
792 | ||
2c8503f5 PNA |
793 | static unsigned int *tcp_get_timeouts(struct net *net) |
794 | { | |
be0593c6 | 795 | return tcp_pernet(net)->timeouts; |
2c8503f5 PNA |
796 | } |
797 | ||
9fb9cbb1 | 798 | /* Returns verdict for packet, or -1 for invalid. */ |
c88130bc | 799 | static int tcp_packet(struct nf_conn *ct, |
9fb9cbb1 YK |
800 | const struct sk_buff *skb, |
801 | unsigned int dataoff, | |
802 | enum ip_conntrack_info ctinfo, | |
76108cea | 803 | u_int8_t pf, |
2c8503f5 PNA |
804 | unsigned int hooknum, |
805 | unsigned int *timeouts) | |
9fb9cbb1 | 806 | { |
c2a2c7e0 | 807 | struct net *net = nf_ct_net(ct); |
d2ba1fde | 808 | struct nf_tcp_net *tn = tcp_pernet(net); |
0d53778e | 809 | struct nf_conntrack_tuple *tuple; |
9fb9cbb1 YK |
810 | enum tcp_conntrack new_state, old_state; |
811 | enum ip_conntrack_dir dir; | |
82f568fc JE |
812 | const struct tcphdr *th; |
813 | struct tcphdr _tcph; | |
9fb9cbb1 YK |
814 | unsigned long timeout; |
815 | unsigned int index; | |
816 | ||
817 | th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph); | |
818 | BUG_ON(th == NULL); | |
819 | ||
440f0d58 | 820 | spin_lock_bh(&ct->lock); |
c88130bc | 821 | old_state = ct->proto.tcp.state; |
9fb9cbb1 YK |
822 | dir = CTINFO2DIR(ctinfo); |
823 | index = get_conntrack_index(th); | |
824 | new_state = tcp_conntracks[dir][index][old_state]; | |
c88130bc | 825 | tuple = &ct->tuplehash[dir].tuple; |
9fb9cbb1 YK |
826 | |
827 | switch (new_state) { | |
17311393 JK |
828 | case TCP_CONNTRACK_SYN_SENT: |
829 | if (old_state < TCP_CONNTRACK_TIME_WAIT) | |
830 | break; | |
b2155e7f JK |
831 | /* RFC 1122: "When a connection is closed actively, |
832 | * it MUST linger in TIME-WAIT state for a time 2xMSL | |
833 | * (Maximum Segment Lifetime). However, it MAY accept | |
834 | * a new SYN from the remote TCP to reopen the connection | |
835 | * directly from TIME-WAIT state, if..." | |
836 | * We ignore the conditions because we are in the | |
837 | * TIME-WAIT state anyway. | |
838 | * | |
839 | * Handle aborted connections: we and the server | |
840 | * think there is an existing connection but the client | |
841 | * aborts it and starts a new one. | |
842 | */ | |
843 | if (((ct->proto.tcp.seen[dir].flags | |
844 | | ct->proto.tcp.seen[!dir].flags) | |
845 | & IP_CT_TCP_FLAG_CLOSE_INIT) | |
c88130bc PM |
846 | || (ct->proto.tcp.last_dir == dir |
847 | && ct->proto.tcp.last_index == TCP_RST_SET)) { | |
bc34b841 JK |
848 | /* Attempt to reopen a closed/aborted connection. |
849 | * Delete this connection and look up again. */ | |
440f0d58 | 850 | spin_unlock_bh(&ct->lock); |
2aec609f | 851 | |
6b69fe0c PM |
852 | /* Only repeat if we can actually remove the timer. |
853 | * Destruction may already be in progress in process | |
854 | * context and we must give it a chance to terminate. | |
855 | */ | |
2aec609f | 856 | if (nf_ct_kill(ct)) |
6b69fe0c | 857 | return -NF_REPEAT; |
ec8d5409 | 858 | return NF_DROP; |
17311393 JK |
859 | } |
860 | /* Fall through */ | |
9fb9cbb1 | 861 | case TCP_CONNTRACK_IGNORE: |
73f30602 | 862 | /* Ignored packets: |
b2155e7f JK |
863 | * |
864 | * Our connection entry may be out of sync, so ignore | |
865 | * packets which may signal the real connection between | |
866 | * the client and the server. | |
73f30602 JK |
867 | * |
868 | * a) SYN in ORIGINAL | |
869 | * b) SYN/ACK in REPLY | |
601e68e1 | 870 | * c) ACK in reply direction after initial SYN in original. |
b2155e7f JK |
871 | * |
872 | * If the ignored packet is invalid, the receiver will send | |
873 | * a RST we'll catch below. | |
73f30602 | 874 | */ |
9fb9cbb1 | 875 | if (index == TCP_SYNACK_SET |
c88130bc PM |
876 | && ct->proto.tcp.last_index == TCP_SYN_SET |
877 | && ct->proto.tcp.last_dir != dir | |
878 | && ntohl(th->ack_seq) == ct->proto.tcp.last_end) { | |
b2155e7f | 879 | /* b) This SYN/ACK acknowledges a SYN that we earlier |
9fb9cbb1 YK |
880 | * ignored as invalid. This means that the client and |
881 | * the server are both in sync, while the firewall is | |
c4832c7b PNA |
882 | * not. We get in sync from the previously annotated |
883 | * values. | |
9fb9cbb1 | 884 | */ |
c4832c7b PNA |
885 | old_state = TCP_CONNTRACK_SYN_SENT; |
886 | new_state = TCP_CONNTRACK_SYN_RECV; | |
887 | ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_end = | |
888 | ct->proto.tcp.last_end; | |
889 | ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxend = | |
890 | ct->proto.tcp.last_end; | |
891 | ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxwin = | |
892 | ct->proto.tcp.last_win == 0 ? | |
893 | 1 : ct->proto.tcp.last_win; | |
894 | ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_scale = | |
895 | ct->proto.tcp.last_wscale; | |
b3cad287 | 896 | ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK; |
c4832c7b PNA |
897 | ct->proto.tcp.seen[ct->proto.tcp.last_dir].flags = |
898 | ct->proto.tcp.last_flags; | |
899 | memset(&ct->proto.tcp.seen[dir], 0, | |
900 | sizeof(struct ip_ct_tcp_state)); | |
901 | break; | |
9fb9cbb1 | 902 | } |
c88130bc PM |
903 | ct->proto.tcp.last_index = index; |
904 | ct->proto.tcp.last_dir = dir; | |
905 | ct->proto.tcp.last_seq = ntohl(th->seq); | |
906 | ct->proto.tcp.last_end = | |
9fb9cbb1 | 907 | segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th); |
c4832c7b PNA |
908 | ct->proto.tcp.last_win = ntohs(th->window); |
909 | ||
910 | /* a) This is a SYN in ORIGINAL. The client and the server | |
911 | * may be in sync but we are not. In that case, we annotate | |
912 | * the TCP options and let the packet go through. If it is a | |
913 | * valid SYN packet, the server will reply with a SYN/ACK, and | |
b3cad287 JDB |
914 | * then we'll get in sync. Otherwise, the server potentially |
915 | * responds with a challenge ACK if implementing RFC5961. | |
916 | */ | |
c4832c7b PNA |
917 | if (index == TCP_SYN_SET && dir == IP_CT_DIR_ORIGINAL) { |
918 | struct ip_ct_tcp_state seen = {}; | |
919 | ||
920 | ct->proto.tcp.last_flags = | |
921 | ct->proto.tcp.last_wscale = 0; | |
922 | tcp_options(skb, dataoff, th, &seen); | |
923 | if (seen.flags & IP_CT_TCP_FLAG_WINDOW_SCALE) { | |
924 | ct->proto.tcp.last_flags |= | |
925 | IP_CT_TCP_FLAG_WINDOW_SCALE; | |
926 | ct->proto.tcp.last_wscale = seen.td_scale; | |
927 | } | |
928 | if (seen.flags & IP_CT_TCP_FLAG_SACK_PERM) { | |
929 | ct->proto.tcp.last_flags |= | |
930 | IP_CT_TCP_FLAG_SACK_PERM; | |
931 | } | |
b3cad287 JDB |
932 | /* Mark the potential for RFC5961 challenge ACK, |
933 | * this pose a special problem for LAST_ACK state | |
934 | * as ACK is intrepretated as ACKing last FIN. | |
935 | */ | |
936 | if (old_state == TCP_CONNTRACK_LAST_ACK) | |
937 | ct->proto.tcp.last_flags |= | |
938 | IP_CT_EXP_CHALLENGE_ACK; | |
c4832c7b | 939 | } |
440f0d58 | 940 | spin_unlock_bh(&ct->lock); |
c2a2c7e0 | 941 | if (LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 | 942 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, |
1a4ac987 PNA |
943 | "nf_ct_tcp: invalid packet ignored in " |
944 | "state %s ", tcp_conntrack_names[old_state]); | |
9fb9cbb1 YK |
945 | return NF_ACCEPT; |
946 | case TCP_CONNTRACK_MAX: | |
48b1de4c PM |
947 | /* Special case for SYN proxy: when the SYN to the server or |
948 | * the SYN/ACK from the server is lost, the client may transmit | |
949 | * a keep-alive packet while in SYN_SENT state. This needs to | |
950 | * be associated with the original conntrack entry in order to | |
951 | * generate a new SYN with the correct sequence number. | |
952 | */ | |
953 | if (nfct_synproxy(ct) && old_state == TCP_CONNTRACK_SYN_SENT && | |
954 | index == TCP_ACK_SET && dir == IP_CT_DIR_ORIGINAL && | |
955 | ct->proto.tcp.last_dir == IP_CT_DIR_ORIGINAL && | |
956 | ct->proto.tcp.seen[dir].td_end - 1 == ntohl(th->seq)) { | |
957 | pr_debug("nf_ct_tcp: SYN proxy client keep alive\n"); | |
958 | spin_unlock_bh(&ct->lock); | |
959 | return NF_ACCEPT; | |
960 | } | |
961 | ||
9fb9cbb1 | 962 | /* Invalid packet */ |
0d53778e PM |
963 | pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n", |
964 | dir, get_conntrack_index(th), old_state); | |
440f0d58 | 965 | spin_unlock_bh(&ct->lock); |
c2a2c7e0 | 966 | if (LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 | 967 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, |
9fb9cbb1 YK |
968 | "nf_ct_tcp: invalid state "); |
969 | return -NF_ACCEPT; | |
b3cad287 JDB |
970 | case TCP_CONNTRACK_TIME_WAIT: |
971 | /* RFC5961 compliance cause stack to send "challenge-ACK" | |
972 | * e.g. in response to spurious SYNs. Conntrack MUST | |
973 | * not believe this ACK is acking last FIN. | |
974 | */ | |
975 | if (old_state == TCP_CONNTRACK_LAST_ACK && | |
976 | index == TCP_ACK_SET && | |
977 | ct->proto.tcp.last_dir != dir && | |
978 | ct->proto.tcp.last_index == TCP_SYN_SET && | |
979 | (ct->proto.tcp.last_flags & IP_CT_EXP_CHALLENGE_ACK)) { | |
980 | /* Detected RFC5961 challenge ACK */ | |
981 | ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK; | |
982 | spin_unlock_bh(&ct->lock); | |
983 | if (LOG_INVALID(net, IPPROTO_TCP)) | |
984 | nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, | |
985 | "nf_ct_tcp: challenge-ACK ignored "); | |
986 | return NF_ACCEPT; /* Don't change state */ | |
987 | } | |
988 | break; | |
9fb9cbb1 | 989 | case TCP_CONNTRACK_CLOSE: |
bfcaa502 JK |
990 | if (index == TCP_RST_SET |
991 | && (ct->proto.tcp.seen[!dir].flags & IP_CT_TCP_FLAG_MAXACK_SET) | |
992 | && before(ntohl(th->seq), ct->proto.tcp.seen[!dir].td_maxack)) { | |
993 | /* Invalid RST */ | |
334a47f6 | 994 | spin_unlock_bh(&ct->lock); |
bfcaa502 | 995 | if (LOG_INVALID(net, IPPROTO_TCP)) |
30e0c6a6 G |
996 | nf_log_packet(net, pf, 0, skb, NULL, NULL, |
997 | NULL, "nf_ct_tcp: invalid RST "); | |
bfcaa502 JK |
998 | return -NF_ACCEPT; |
999 | } | |
9fb9cbb1 | 1000 | if (index == TCP_RST_SET |
c88130bc PM |
1001 | && ((test_bit(IPS_SEEN_REPLY_BIT, &ct->status) |
1002 | && ct->proto.tcp.last_index == TCP_SYN_SET) | |
1003 | || (!test_bit(IPS_ASSURED_BIT, &ct->status) | |
1004 | && ct->proto.tcp.last_index == TCP_ACK_SET)) | |
1005 | && ntohl(th->ack_seq) == ct->proto.tcp.last_end) { | |
93b1fae4 | 1006 | /* RST sent to invalid SYN or ACK we had let through |
73f30602 JK |
1007 | * at a) and c) above: |
1008 | * | |
1009 | * a) SYN was in window then | |
1010 | * c) we hold a half-open connection. | |
1011 | * | |
1012 | * Delete our connection entry. | |
9fb9cbb1 | 1013 | * We skip window checking, because packet might ACK |
73f30602 | 1014 | * segments we ignored. */ |
9fb9cbb1 YK |
1015 | goto in_window; |
1016 | } | |
93b1fae4 | 1017 | /* Just fall through */ |
9fb9cbb1 YK |
1018 | default: |
1019 | /* Keep compilers happy. */ | |
1020 | break; | |
1021 | } | |
1022 | ||
c88130bc | 1023 | if (!tcp_in_window(ct, &ct->proto.tcp, dir, index, |
9fb9cbb1 | 1024 | skb, dataoff, th, pf)) { |
440f0d58 | 1025 | spin_unlock_bh(&ct->lock); |
9fb9cbb1 YK |
1026 | return -NF_ACCEPT; |
1027 | } | |
1028 | in_window: | |
1029 | /* From now on we have got in-window packets */ | |
c88130bc PM |
1030 | ct->proto.tcp.last_index = index; |
1031 | ct->proto.tcp.last_dir = dir; | |
9fb9cbb1 | 1032 | |
0d53778e | 1033 | pr_debug("tcp_conntracks: "); |
3c9fba65 | 1034 | nf_ct_dump_tuple(tuple); |
0d53778e PM |
1035 | pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n", |
1036 | (th->syn ? 1 : 0), (th->ack ? 1 : 0), | |
1037 | (th->fin ? 1 : 0), (th->rst ? 1 : 0), | |
1038 | old_state, new_state); | |
9fb9cbb1 | 1039 | |
c88130bc | 1040 | ct->proto.tcp.state = new_state; |
9fb9cbb1 | 1041 | if (old_state != new_state |
d0c1fd7a | 1042 | && new_state == TCP_CONNTRACK_FIN_WAIT) |
c88130bc | 1043 | ct->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT; |
ae375044 | 1044 | |
d2ba1fde | 1045 | if (ct->proto.tcp.retrans >= tn->tcp_max_retrans && |
2c8503f5 PNA |
1046 | timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS]) |
1047 | timeout = timeouts[TCP_CONNTRACK_RETRANS]; | |
ae375044 PM |
1048 | else if ((ct->proto.tcp.seen[0].flags | ct->proto.tcp.seen[1].flags) & |
1049 | IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED && | |
2c8503f5 PNA |
1050 | timeouts[new_state] > timeouts[TCP_CONNTRACK_UNACK]) |
1051 | timeout = timeouts[TCP_CONNTRACK_UNACK]; | |
ae375044 | 1052 | else |
2c8503f5 | 1053 | timeout = timeouts[new_state]; |
440f0d58 | 1054 | spin_unlock_bh(&ct->lock); |
9fb9cbb1 | 1055 | |
9fb9cbb1 | 1056 | if (new_state != old_state) |
a71996fc | 1057 | nf_conntrack_event_cache(IPCT_PROTOINFO, ct); |
9fb9cbb1 | 1058 | |
c88130bc | 1059 | if (!test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { |
9fb9cbb1 YK |
1060 | /* If only reply is a RST, we can consider ourselves not to |
1061 | have an established connection: this is a fairly common | |
1062 | problem case, so we can delete the conntrack | |
1063 | immediately. --RR */ | |
1064 | if (th->rst) { | |
718d4ad9 | 1065 | nf_ct_kill_acct(ct, ctinfo, skb); |
9fb9cbb1 YK |
1066 | return NF_ACCEPT; |
1067 | } | |
6547a221 FW |
1068 | /* ESTABLISHED without SEEN_REPLY, i.e. mid-connection |
1069 | * pickup with loose=1. Avoid large ESTABLISHED timeout. | |
1070 | */ | |
1071 | if (new_state == TCP_CONNTRACK_ESTABLISHED && | |
1072 | timeout > timeouts[TCP_CONNTRACK_UNACK]) | |
1073 | timeout = timeouts[TCP_CONNTRACK_UNACK]; | |
c88130bc | 1074 | } else if (!test_bit(IPS_ASSURED_BIT, &ct->status) |
9fb9cbb1 YK |
1075 | && (old_state == TCP_CONNTRACK_SYN_RECV |
1076 | || old_state == TCP_CONNTRACK_ESTABLISHED) | |
1077 | && new_state == TCP_CONNTRACK_ESTABLISHED) { | |
601e68e1 YH |
1078 | /* Set ASSURED if we see see valid ack in ESTABLISHED |
1079 | after SYN_RECV or a valid answer for a picked up | |
9fb9cbb1 | 1080 | connection. */ |
c88130bc | 1081 | set_bit(IPS_ASSURED_BIT, &ct->status); |
858b3133 | 1082 | nf_conntrack_event_cache(IPCT_ASSURED, ct); |
9fb9cbb1 | 1083 | } |
c88130bc | 1084 | nf_ct_refresh_acct(ct, ctinfo, skb, timeout); |
9fb9cbb1 YK |
1085 | |
1086 | return NF_ACCEPT; | |
1087 | } | |
601e68e1 | 1088 | |
9fb9cbb1 | 1089 | /* Called when a new connection for this protocol found. */ |
09f263cd | 1090 | static bool tcp_new(struct nf_conn *ct, const struct sk_buff *skb, |
2c8503f5 | 1091 | unsigned int dataoff, unsigned int *timeouts) |
9fb9cbb1 YK |
1092 | { |
1093 | enum tcp_conntrack new_state; | |
82f568fc JE |
1094 | const struct tcphdr *th; |
1095 | struct tcphdr _tcph; | |
d2ba1fde G |
1096 | struct net *net = nf_ct_net(ct); |
1097 | struct nf_tcp_net *tn = tcp_pernet(net); | |
82f568fc JE |
1098 | const struct ip_ct_tcp_state *sender = &ct->proto.tcp.seen[0]; |
1099 | const struct ip_ct_tcp_state *receiver = &ct->proto.tcp.seen[1]; | |
9fb9cbb1 YK |
1100 | |
1101 | th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph); | |
1102 | BUG_ON(th == NULL); | |
1103 | ||
1104 | /* Don't need lock here: this conntrack not in circulation yet */ | |
e5fc9e7a | 1105 | new_state = tcp_conntracks[0][get_conntrack_index(th)][TCP_CONNTRACK_NONE]; |
9fb9cbb1 YK |
1106 | |
1107 | /* Invalid: delete conntrack */ | |
1108 | if (new_state >= TCP_CONNTRACK_MAX) { | |
0d53778e | 1109 | pr_debug("nf_ct_tcp: invalid new deleting.\n"); |
09f263cd | 1110 | return false; |
9fb9cbb1 YK |
1111 | } |
1112 | ||
1113 | if (new_state == TCP_CONNTRACK_SYN_SENT) { | |
e5fc9e7a | 1114 | memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp)); |
9fb9cbb1 | 1115 | /* SYN packet */ |
c88130bc | 1116 | ct->proto.tcp.seen[0].td_end = |
9fb9cbb1 YK |
1117 | segment_seq_plus_len(ntohl(th->seq), skb->len, |
1118 | dataoff, th); | |
c88130bc PM |
1119 | ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window); |
1120 | if (ct->proto.tcp.seen[0].td_maxwin == 0) | |
1121 | ct->proto.tcp.seen[0].td_maxwin = 1; | |
1122 | ct->proto.tcp.seen[0].td_maxend = | |
1123 | ct->proto.tcp.seen[0].td_end; | |
1124 | ||
1125 | tcp_options(skb, dataoff, th, &ct->proto.tcp.seen[0]); | |
d2ba1fde | 1126 | } else if (tn->tcp_loose == 0) { |
9fb9cbb1 | 1127 | /* Don't try to pick up connections. */ |
09f263cd | 1128 | return false; |
9fb9cbb1 | 1129 | } else { |
e5fc9e7a | 1130 | memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp)); |
9fb9cbb1 YK |
1131 | /* |
1132 | * We are in the middle of a connection, | |
1133 | * its history is lost for us. | |
1134 | * Let's try to use the data from the packet. | |
1135 | */ | |
c88130bc | 1136 | ct->proto.tcp.seen[0].td_end = |
9fb9cbb1 YK |
1137 | segment_seq_plus_len(ntohl(th->seq), skb->len, |
1138 | dataoff, th); | |
c88130bc PM |
1139 | ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window); |
1140 | if (ct->proto.tcp.seen[0].td_maxwin == 0) | |
1141 | ct->proto.tcp.seen[0].td_maxwin = 1; | |
1142 | ct->proto.tcp.seen[0].td_maxend = | |
1143 | ct->proto.tcp.seen[0].td_end + | |
1144 | ct->proto.tcp.seen[0].td_maxwin; | |
9fb9cbb1 | 1145 | |
a09113c2 PM |
1146 | /* We assume SACK and liberal window checking to handle |
1147 | * window scaling */ | |
c88130bc PM |
1148 | ct->proto.tcp.seen[0].flags = |
1149 | ct->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM | | |
1150 | IP_CT_TCP_FLAG_BE_LIBERAL; | |
9fb9cbb1 | 1151 | } |
601e68e1 | 1152 | |
9fb9cbb1 | 1153 | /* tcp_packet will set them */ |
c88130bc | 1154 | ct->proto.tcp.last_index = TCP_NONE_SET; |
601e68e1 | 1155 | |
0d53778e PM |
1156 | pr_debug("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i " |
1157 | "receiver end=%u maxend=%u maxwin=%u scale=%i\n", | |
1158 | sender->td_end, sender->td_maxend, sender->td_maxwin, | |
1159 | sender->td_scale, | |
1160 | receiver->td_end, receiver->td_maxend, receiver->td_maxwin, | |
1161 | receiver->td_scale); | |
09f263cd | 1162 | return true; |
9fb9cbb1 | 1163 | } |
c1d10adb | 1164 | |
c6dd940b FW |
1165 | static bool tcp_can_early_drop(const struct nf_conn *ct) |
1166 | { | |
1167 | switch (ct->proto.tcp.state) { | |
1168 | case TCP_CONNTRACK_FIN_WAIT: | |
1169 | case TCP_CONNTRACK_LAST_ACK: | |
1170 | case TCP_CONNTRACK_TIME_WAIT: | |
1171 | case TCP_CONNTRACK_CLOSE: | |
1172 | case TCP_CONNTRACK_CLOSE_WAIT: | |
1173 | return true; | |
1174 | default: | |
1175 | break; | |
1176 | } | |
1177 | ||
1178 | return false; | |
1179 | } | |
1180 | ||
c0cd1156 | 1181 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
c1d10adb PNA |
1182 | |
1183 | #include <linux/netfilter/nfnetlink.h> | |
1184 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
1185 | ||
fdf70832 | 1186 | static int tcp_to_nlattr(struct sk_buff *skb, struct nlattr *nla, |
440f0d58 | 1187 | struct nf_conn *ct) |
c1d10adb | 1188 | { |
df6fb868 | 1189 | struct nlattr *nest_parms; |
c8e2078c | 1190 | struct nf_ct_tcp_flags tmp = {}; |
601e68e1 | 1191 | |
440f0d58 | 1192 | spin_lock_bh(&ct->lock); |
df6fb868 PM |
1193 | nest_parms = nla_nest_start(skb, CTA_PROTOINFO_TCP | NLA_F_NESTED); |
1194 | if (!nest_parms) | |
1195 | goto nla_put_failure; | |
1196 | ||
4925a459 DM |
1197 | if (nla_put_u8(skb, CTA_PROTOINFO_TCP_STATE, ct->proto.tcp.state) || |
1198 | nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL, | |
1199 | ct->proto.tcp.seen[0].td_scale) || | |
1200 | nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY, | |
1201 | ct->proto.tcp.seen[1].td_scale)) | |
1202 | goto nla_put_failure; | |
c8e2078c PNA |
1203 | |
1204 | tmp.flags = ct->proto.tcp.seen[0].flags; | |
4925a459 DM |
1205 | if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL, |
1206 | sizeof(struct nf_ct_tcp_flags), &tmp)) | |
1207 | goto nla_put_failure; | |
c8e2078c PNA |
1208 | |
1209 | tmp.flags = ct->proto.tcp.seen[1].flags; | |
4925a459 DM |
1210 | if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY, |
1211 | sizeof(struct nf_ct_tcp_flags), &tmp)) | |
1212 | goto nla_put_failure; | |
440f0d58 | 1213 | spin_unlock_bh(&ct->lock); |
c1d10adb | 1214 | |
df6fb868 | 1215 | nla_nest_end(skb, nest_parms); |
c1d10adb PNA |
1216 | |
1217 | return 0; | |
1218 | ||
df6fb868 | 1219 | nla_put_failure: |
440f0d58 | 1220 | spin_unlock_bh(&ct->lock); |
c1d10adb PNA |
1221 | return -1; |
1222 | } | |
1223 | ||
f73e924c PM |
1224 | static const struct nla_policy tcp_nla_policy[CTA_PROTOINFO_TCP_MAX+1] = { |
1225 | [CTA_PROTOINFO_TCP_STATE] = { .type = NLA_U8 }, | |
1226 | [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] = { .type = NLA_U8 }, | |
1227 | [CTA_PROTOINFO_TCP_WSCALE_REPLY] = { .type = NLA_U8 }, | |
1228 | [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL] = { .len = sizeof(struct nf_ct_tcp_flags) }, | |
1229 | [CTA_PROTOINFO_TCP_FLAGS_REPLY] = { .len = sizeof(struct nf_ct_tcp_flags) }, | |
c1d10adb PNA |
1230 | }; |
1231 | ||
fdf70832 | 1232 | static int nlattr_to_tcp(struct nlattr *cda[], struct nf_conn *ct) |
c1d10adb | 1233 | { |
2f0d2f10 | 1234 | struct nlattr *pattr = cda[CTA_PROTOINFO_TCP]; |
df6fb868 | 1235 | struct nlattr *tb[CTA_PROTOINFO_TCP_MAX+1]; |
f73e924c | 1236 | int err; |
c1d10adb PNA |
1237 | |
1238 | /* updates could not contain anything about the private | |
1239 | * protocol info, in that case skip the parsing */ | |
2f0d2f10 | 1240 | if (!pattr) |
c1d10adb PNA |
1241 | return 0; |
1242 | ||
fceb6435 JB |
1243 | err = nla_parse_nested(tb, CTA_PROTOINFO_TCP_MAX, pattr, |
1244 | tcp_nla_policy, NULL); | |
f73e924c PM |
1245 | if (err < 0) |
1246 | return err; | |
c1d10adb | 1247 | |
5f7da4d2 PM |
1248 | if (tb[CTA_PROTOINFO_TCP_STATE] && |
1249 | nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]) >= TCP_CONNTRACK_MAX) | |
c1d10adb PNA |
1250 | return -EINVAL; |
1251 | ||
440f0d58 | 1252 | spin_lock_bh(&ct->lock); |
5f7da4d2 PM |
1253 | if (tb[CTA_PROTOINFO_TCP_STATE]) |
1254 | ct->proto.tcp.state = nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]); | |
c8e2078c | 1255 | |
df6fb868 | 1256 | if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]) { |
c8e2078c | 1257 | struct nf_ct_tcp_flags *attr = |
df6fb868 | 1258 | nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]); |
c8e2078c PNA |
1259 | ct->proto.tcp.seen[0].flags &= ~attr->mask; |
1260 | ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask; | |
1261 | } | |
1262 | ||
df6fb868 | 1263 | if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]) { |
c8e2078c | 1264 | struct nf_ct_tcp_flags *attr = |
df6fb868 | 1265 | nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]); |
c8e2078c PNA |
1266 | ct->proto.tcp.seen[1].flags &= ~attr->mask; |
1267 | ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask; | |
1268 | } | |
1269 | ||
df6fb868 PM |
1270 | if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] && |
1271 | tb[CTA_PROTOINFO_TCP_WSCALE_REPLY] && | |
c8e2078c PNA |
1272 | ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_WINDOW_SCALE && |
1273 | ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_WINDOW_SCALE) { | |
77236b6e PM |
1274 | ct->proto.tcp.seen[0].td_scale = |
1275 | nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL]); | |
1276 | ct->proto.tcp.seen[1].td_scale = | |
1277 | nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_REPLY]); | |
c8e2078c | 1278 | } |
440f0d58 | 1279 | spin_unlock_bh(&ct->lock); |
c1d10adb PNA |
1280 | |
1281 | return 0; | |
1282 | } | |
a400c30e HE |
1283 | |
1284 | static int tcp_nlattr_size(void) | |
1285 | { | |
1286 | return nla_total_size(0) /* CTA_PROTOINFO_TCP */ | |
1287 | + nla_policy_len(tcp_nla_policy, CTA_PROTOINFO_TCP_MAX + 1); | |
1288 | } | |
1289 | ||
1290 | static int tcp_nlattr_tuple_size(void) | |
1291 | { | |
1292 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
1293 | } | |
c1d10adb | 1294 | #endif |
933a41e7 | 1295 | |
50978462 PNA |
1296 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) |
1297 | ||
1298 | #include <linux/netfilter/nfnetlink.h> | |
1299 | #include <linux/netfilter/nfnetlink_cttimeout.h> | |
1300 | ||
8264deb8 G |
1301 | static int tcp_timeout_nlattr_to_obj(struct nlattr *tb[], |
1302 | struct net *net, void *data) | |
50978462 PNA |
1303 | { |
1304 | unsigned int *timeouts = data; | |
8264deb8 | 1305 | struct nf_tcp_net *tn = tcp_pernet(net); |
50978462 PNA |
1306 | int i; |
1307 | ||
1308 | /* set default TCP timeouts. */ | |
1309 | for (i=0; i<TCP_CONNTRACK_TIMEOUT_MAX; i++) | |
8264deb8 | 1310 | timeouts[i] = tn->timeouts[i]; |
50978462 PNA |
1311 | |
1312 | if (tb[CTA_TIMEOUT_TCP_SYN_SENT]) { | |
1313 | timeouts[TCP_CONNTRACK_SYN_SENT] = | |
1314 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT]))*HZ; | |
1315 | } | |
1316 | if (tb[CTA_TIMEOUT_TCP_SYN_RECV]) { | |
1317 | timeouts[TCP_CONNTRACK_SYN_RECV] = | |
1318 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_RECV]))*HZ; | |
1319 | } | |
1320 | if (tb[CTA_TIMEOUT_TCP_ESTABLISHED]) { | |
1321 | timeouts[TCP_CONNTRACK_ESTABLISHED] = | |
1322 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_ESTABLISHED]))*HZ; | |
1323 | } | |
1324 | if (tb[CTA_TIMEOUT_TCP_FIN_WAIT]) { | |
1325 | timeouts[TCP_CONNTRACK_FIN_WAIT] = | |
1326 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_FIN_WAIT]))*HZ; | |
1327 | } | |
1328 | if (tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]) { | |
1329 | timeouts[TCP_CONNTRACK_CLOSE_WAIT] = | |
1330 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]))*HZ; | |
1331 | } | |
1332 | if (tb[CTA_TIMEOUT_TCP_LAST_ACK]) { | |
1333 | timeouts[TCP_CONNTRACK_LAST_ACK] = | |
1334 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_LAST_ACK]))*HZ; | |
1335 | } | |
1336 | if (tb[CTA_TIMEOUT_TCP_TIME_WAIT]) { | |
1337 | timeouts[TCP_CONNTRACK_TIME_WAIT] = | |
1338 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_TIME_WAIT]))*HZ; | |
1339 | } | |
1340 | if (tb[CTA_TIMEOUT_TCP_CLOSE]) { | |
1341 | timeouts[TCP_CONNTRACK_CLOSE] = | |
1342 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE]))*HZ; | |
1343 | } | |
1344 | if (tb[CTA_TIMEOUT_TCP_SYN_SENT2]) { | |
1345 | timeouts[TCP_CONNTRACK_SYN_SENT2] = | |
1346 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT2]))*HZ; | |
1347 | } | |
1348 | if (tb[CTA_TIMEOUT_TCP_RETRANS]) { | |
1349 | timeouts[TCP_CONNTRACK_RETRANS] = | |
1350 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_RETRANS]))*HZ; | |
1351 | } | |
1352 | if (tb[CTA_TIMEOUT_TCP_UNACK]) { | |
1353 | timeouts[TCP_CONNTRACK_UNACK] = | |
1354 | ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_UNACK]))*HZ; | |
1355 | } | |
1356 | return 0; | |
1357 | } | |
1358 | ||
1359 | static int | |
1360 | tcp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data) | |
1361 | { | |
1362 | const unsigned int *timeouts = data; | |
1363 | ||
4925a459 DM |
1364 | if (nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT, |
1365 | htonl(timeouts[TCP_CONNTRACK_SYN_SENT] / HZ)) || | |
1366 | nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_RECV, | |
1367 | htonl(timeouts[TCP_CONNTRACK_SYN_RECV] / HZ)) || | |
1368 | nla_put_be32(skb, CTA_TIMEOUT_TCP_ESTABLISHED, | |
1369 | htonl(timeouts[TCP_CONNTRACK_ESTABLISHED] / HZ)) || | |
1370 | nla_put_be32(skb, CTA_TIMEOUT_TCP_FIN_WAIT, | |
1371 | htonl(timeouts[TCP_CONNTRACK_FIN_WAIT] / HZ)) || | |
1372 | nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE_WAIT, | |
1373 | htonl(timeouts[TCP_CONNTRACK_CLOSE_WAIT] / HZ)) || | |
1374 | nla_put_be32(skb, CTA_TIMEOUT_TCP_LAST_ACK, | |
1375 | htonl(timeouts[TCP_CONNTRACK_LAST_ACK] / HZ)) || | |
1376 | nla_put_be32(skb, CTA_TIMEOUT_TCP_TIME_WAIT, | |
1377 | htonl(timeouts[TCP_CONNTRACK_TIME_WAIT] / HZ)) || | |
1378 | nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE, | |
1379 | htonl(timeouts[TCP_CONNTRACK_CLOSE] / HZ)) || | |
1380 | nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT2, | |
1381 | htonl(timeouts[TCP_CONNTRACK_SYN_SENT2] / HZ)) || | |
1382 | nla_put_be32(skb, CTA_TIMEOUT_TCP_RETRANS, | |
1383 | htonl(timeouts[TCP_CONNTRACK_RETRANS] / HZ)) || | |
1384 | nla_put_be32(skb, CTA_TIMEOUT_TCP_UNACK, | |
1385 | htonl(timeouts[TCP_CONNTRACK_UNACK] / HZ))) | |
1386 | goto nla_put_failure; | |
50978462 PNA |
1387 | return 0; |
1388 | ||
1389 | nla_put_failure: | |
1390 | return -ENOSPC; | |
1391 | } | |
1392 | ||
1393 | static const struct nla_policy tcp_timeout_nla_policy[CTA_TIMEOUT_TCP_MAX+1] = { | |
1394 | [CTA_TIMEOUT_TCP_SYN_SENT] = { .type = NLA_U32 }, | |
1395 | [CTA_TIMEOUT_TCP_SYN_RECV] = { .type = NLA_U32 }, | |
1396 | [CTA_TIMEOUT_TCP_ESTABLISHED] = { .type = NLA_U32 }, | |
1397 | [CTA_TIMEOUT_TCP_FIN_WAIT] = { .type = NLA_U32 }, | |
1398 | [CTA_TIMEOUT_TCP_CLOSE_WAIT] = { .type = NLA_U32 }, | |
1399 | [CTA_TIMEOUT_TCP_LAST_ACK] = { .type = NLA_U32 }, | |
1400 | [CTA_TIMEOUT_TCP_TIME_WAIT] = { .type = NLA_U32 }, | |
1401 | [CTA_TIMEOUT_TCP_CLOSE] = { .type = NLA_U32 }, | |
1402 | [CTA_TIMEOUT_TCP_SYN_SENT2] = { .type = NLA_U32 }, | |
6d1fafca FW |
1403 | [CTA_TIMEOUT_TCP_RETRANS] = { .type = NLA_U32 }, |
1404 | [CTA_TIMEOUT_TCP_UNACK] = { .type = NLA_U32 }, | |
50978462 PNA |
1405 | }; |
1406 | #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */ | |
1407 | ||
933a41e7 | 1408 | #ifdef CONFIG_SYSCTL |
933a41e7 PM |
1409 | static struct ctl_table tcp_sysctl_table[] = { |
1410 | { | |
933a41e7 | 1411 | .procname = "nf_conntrack_tcp_timeout_syn_sent", |
933a41e7 PM |
1412 | .maxlen = sizeof(unsigned int), |
1413 | .mode = 0644, | |
6d9f239a | 1414 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1415 | }, |
1416 | { | |
933a41e7 | 1417 | .procname = "nf_conntrack_tcp_timeout_syn_recv", |
933a41e7 PM |
1418 | .maxlen = sizeof(unsigned int), |
1419 | .mode = 0644, | |
6d9f239a | 1420 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1421 | }, |
1422 | { | |
933a41e7 | 1423 | .procname = "nf_conntrack_tcp_timeout_established", |
933a41e7 PM |
1424 | .maxlen = sizeof(unsigned int), |
1425 | .mode = 0644, | |
6d9f239a | 1426 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1427 | }, |
1428 | { | |
933a41e7 | 1429 | .procname = "nf_conntrack_tcp_timeout_fin_wait", |
933a41e7 PM |
1430 | .maxlen = sizeof(unsigned int), |
1431 | .mode = 0644, | |
6d9f239a | 1432 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1433 | }, |
1434 | { | |
933a41e7 | 1435 | .procname = "nf_conntrack_tcp_timeout_close_wait", |
933a41e7 PM |
1436 | .maxlen = sizeof(unsigned int), |
1437 | .mode = 0644, | |
6d9f239a | 1438 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1439 | }, |
1440 | { | |
933a41e7 | 1441 | .procname = "nf_conntrack_tcp_timeout_last_ack", |
933a41e7 PM |
1442 | .maxlen = sizeof(unsigned int), |
1443 | .mode = 0644, | |
6d9f239a | 1444 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1445 | }, |
1446 | { | |
933a41e7 | 1447 | .procname = "nf_conntrack_tcp_timeout_time_wait", |
933a41e7 PM |
1448 | .maxlen = sizeof(unsigned int), |
1449 | .mode = 0644, | |
6d9f239a | 1450 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1451 | }, |
1452 | { | |
933a41e7 | 1453 | .procname = "nf_conntrack_tcp_timeout_close", |
933a41e7 PM |
1454 | .maxlen = sizeof(unsigned int), |
1455 | .mode = 0644, | |
6d9f239a | 1456 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 PM |
1457 | }, |
1458 | { | |
933a41e7 | 1459 | .procname = "nf_conntrack_tcp_timeout_max_retrans", |
933a41e7 PM |
1460 | .maxlen = sizeof(unsigned int), |
1461 | .mode = 0644, | |
6d9f239a | 1462 | .proc_handler = proc_dointvec_jiffies, |
933a41e7 | 1463 | }, |
ae375044 PM |
1464 | { |
1465 | .procname = "nf_conntrack_tcp_timeout_unacknowledged", | |
ae375044 PM |
1466 | .maxlen = sizeof(unsigned int), |
1467 | .mode = 0644, | |
6d9f239a | 1468 | .proc_handler = proc_dointvec_jiffies, |
ae375044 | 1469 | }, |
933a41e7 | 1470 | { |
933a41e7 | 1471 | .procname = "nf_conntrack_tcp_loose", |
933a41e7 PM |
1472 | .maxlen = sizeof(unsigned int), |
1473 | .mode = 0644, | |
6d9f239a | 1474 | .proc_handler = proc_dointvec, |
933a41e7 PM |
1475 | }, |
1476 | { | |
933a41e7 | 1477 | .procname = "nf_conntrack_tcp_be_liberal", |
933a41e7 PM |
1478 | .maxlen = sizeof(unsigned int), |
1479 | .mode = 0644, | |
6d9f239a | 1480 | .proc_handler = proc_dointvec, |
933a41e7 PM |
1481 | }, |
1482 | { | |
933a41e7 | 1483 | .procname = "nf_conntrack_tcp_max_retrans", |
933a41e7 PM |
1484 | .maxlen = sizeof(unsigned int), |
1485 | .mode = 0644, | |
6d9f239a | 1486 | .proc_handler = proc_dointvec, |
933a41e7 | 1487 | }, |
f8572d8f | 1488 | { } |
933a41e7 PM |
1489 | }; |
1490 | #endif /* CONFIG_SYSCTL */ | |
1491 | ||
efa758fe G |
1492 | static int tcp_kmemdup_sysctl_table(struct nf_proto_net *pn, |
1493 | struct nf_tcp_net *tn) | |
d2ba1fde G |
1494 | { |
1495 | #ifdef CONFIG_SYSCTL | |
d2ba1fde G |
1496 | if (pn->ctl_table) |
1497 | return 0; | |
1498 | ||
1499 | pn->ctl_table = kmemdup(tcp_sysctl_table, | |
1500 | sizeof(tcp_sysctl_table), | |
1501 | GFP_KERNEL); | |
1502 | if (!pn->ctl_table) | |
1503 | return -ENOMEM; | |
1504 | ||
1505 | pn->ctl_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT]; | |
1506 | pn->ctl_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV]; | |
1507 | pn->ctl_table[2].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED]; | |
1508 | pn->ctl_table[3].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT]; | |
1509 | pn->ctl_table[4].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT]; | |
1510 | pn->ctl_table[5].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK]; | |
1511 | pn->ctl_table[6].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT]; | |
1512 | pn->ctl_table[7].data = &tn->timeouts[TCP_CONNTRACK_CLOSE]; | |
1513 | pn->ctl_table[8].data = &tn->timeouts[TCP_CONNTRACK_RETRANS]; | |
1514 | pn->ctl_table[9].data = &tn->timeouts[TCP_CONNTRACK_UNACK]; | |
1515 | pn->ctl_table[10].data = &tn->tcp_loose; | |
1516 | pn->ctl_table[11].data = &tn->tcp_be_liberal; | |
1517 | pn->ctl_table[12].data = &tn->tcp_max_retrans; | |
1518 | #endif | |
1519 | return 0; | |
1520 | } | |
1521 | ||
efa758fe | 1522 | static int tcp_init_net(struct net *net, u_int16_t proto) |
d2ba1fde | 1523 | { |
d2ba1fde | 1524 | struct nf_tcp_net *tn = tcp_pernet(net); |
efa758fe G |
1525 | struct nf_proto_net *pn = &tn->pn; |
1526 | ||
1527 | if (!pn->users) { | |
1528 | int i; | |
d2ba1fde | 1529 | |
d2ba1fde G |
1530 | for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++) |
1531 | tn->timeouts[i] = tcp_timeouts[i]; | |
1532 | ||
1533 | tn->tcp_loose = nf_ct_tcp_loose; | |
1534 | tn->tcp_be_liberal = nf_ct_tcp_be_liberal; | |
1535 | tn->tcp_max_retrans = nf_ct_tcp_max_retrans; | |
1536 | } | |
1537 | ||
adf05168 | 1538 | return tcp_kmemdup_sysctl_table(pn, tn); |
d2ba1fde G |
1539 | } |
1540 | ||
08911475 PNA |
1541 | static struct nf_proto_net *tcp_get_net_proto(struct net *net) |
1542 | { | |
1543 | return &net->ct.nf_ct_proto.tcp.pn; | |
1544 | } | |
1545 | ||
61075af5 | 1546 | struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp4 __read_mostly = |
9fb9cbb1 YK |
1547 | { |
1548 | .l3proto = PF_INET, | |
605dcad6 | 1549 | .l4proto = IPPROTO_TCP, |
9fb9cbb1 YK |
1550 | .pkt_to_tuple = tcp_pkt_to_tuple, |
1551 | .invert_tuple = tcp_invert_tuple, | |
9fb9cbb1 YK |
1552 | .print_conntrack = tcp_print_conntrack, |
1553 | .packet = tcp_packet, | |
2c8503f5 | 1554 | .get_timeouts = tcp_get_timeouts, |
9fb9cbb1 | 1555 | .new = tcp_new, |
96f6bf82 | 1556 | .error = tcp_error, |
c6dd940b | 1557 | .can_early_drop = tcp_can_early_drop, |
c0cd1156 | 1558 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
fdf70832 | 1559 | .to_nlattr = tcp_to_nlattr, |
a400c30e | 1560 | .nlattr_size = tcp_nlattr_size, |
fdf70832 PM |
1561 | .from_nlattr = nlattr_to_tcp, |
1562 | .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr, | |
1563 | .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple, | |
a400c30e | 1564 | .nlattr_tuple_size = tcp_nlattr_tuple_size, |
f73e924c | 1565 | .nla_policy = nf_ct_port_nla_policy, |
c1d10adb | 1566 | #endif |
50978462 PNA |
1567 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) |
1568 | .ctnl_timeout = { | |
1569 | .nlattr_to_obj = tcp_timeout_nlattr_to_obj, | |
1570 | .obj_to_nlattr = tcp_timeout_obj_to_nlattr, | |
1571 | .nlattr_max = CTA_TIMEOUT_TCP_MAX, | |
1572 | .obj_size = sizeof(unsigned int) * | |
1573 | TCP_CONNTRACK_TIMEOUT_MAX, | |
1574 | .nla_policy = tcp_timeout_nla_policy, | |
1575 | }, | |
1576 | #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */ | |
efa758fe | 1577 | .init_net = tcp_init_net, |
08911475 | 1578 | .get_net_proto = tcp_get_net_proto, |
9fb9cbb1 | 1579 | }; |
13b18339 | 1580 | EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp4); |
9fb9cbb1 | 1581 | |
61075af5 | 1582 | struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6 __read_mostly = |
9fb9cbb1 YK |
1583 | { |
1584 | .l3proto = PF_INET6, | |
605dcad6 | 1585 | .l4proto = IPPROTO_TCP, |
9fb9cbb1 YK |
1586 | .pkt_to_tuple = tcp_pkt_to_tuple, |
1587 | .invert_tuple = tcp_invert_tuple, | |
9fb9cbb1 YK |
1588 | .print_conntrack = tcp_print_conntrack, |
1589 | .packet = tcp_packet, | |
2c8503f5 | 1590 | .get_timeouts = tcp_get_timeouts, |
9fb9cbb1 | 1591 | .new = tcp_new, |
96f6bf82 | 1592 | .error = tcp_error, |
c6dd940b | 1593 | .can_early_drop = tcp_can_early_drop, |
c0cd1156 | 1594 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
fdf70832 | 1595 | .to_nlattr = tcp_to_nlattr, |
a400c30e | 1596 | .nlattr_size = tcp_nlattr_size, |
fdf70832 PM |
1597 | .from_nlattr = nlattr_to_tcp, |
1598 | .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr, | |
1599 | .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple, | |
a400c30e | 1600 | .nlattr_tuple_size = tcp_nlattr_tuple_size, |
f73e924c | 1601 | .nla_policy = nf_ct_port_nla_policy, |
c1d10adb | 1602 | #endif |
50978462 PNA |
1603 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT) |
1604 | .ctnl_timeout = { | |
1605 | .nlattr_to_obj = tcp_timeout_nlattr_to_obj, | |
1606 | .obj_to_nlattr = tcp_timeout_obj_to_nlattr, | |
1607 | .nlattr_max = CTA_TIMEOUT_TCP_MAX, | |
1608 | .obj_size = sizeof(unsigned int) * | |
1609 | TCP_CONNTRACK_TIMEOUT_MAX, | |
1610 | .nla_policy = tcp_timeout_nla_policy, | |
1611 | }, | |
1612 | #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */ | |
efa758fe | 1613 | .init_net = tcp_init_net, |
08911475 | 1614 | .get_net_proto = tcp_get_net_proto, |
9fb9cbb1 | 1615 | }; |
13b18339 | 1616 | EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp6); |