1 /* (C) 1999-2001 Paul `Rusty' Russell
2 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
3 * (C) 2002-2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
4 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
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.
11 #include <linux/types.h>
12 #include <linux/timer.h>
13 #include <linux/module.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>
20 #include <asm/unaligned.h>
24 #include <linux/netfilter.h>
25 #include <linux/netfilter_ipv4.h>
26 #include <linux/netfilter_ipv6.h>
27 #include <net/netfilter/nf_conntrack.h>
28 #include <net/netfilter/nf_conntrack_l4proto.h>
29 #include <net/netfilter/nf_conntrack_ecache.h>
30 #include <net/netfilter/nf_conntrack_seqadj.h>
31 #include <net/netfilter/nf_conntrack_synproxy.h>
32 #include <net/netfilter/nf_log.h>
33 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
34 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
36 /* "Be conservative in what you do,
37 be liberal in what you accept from others."
38 If it's non-zero, we mark only out of window RST segments as INVALID. */
39 static int nf_ct_tcp_be_liberal __read_mostly
= 0;
41 /* If it is set to zero, we disable picking up already established
43 static int nf_ct_tcp_loose __read_mostly
= 1;
45 /* Max number of the retransmitted packets without receiving an (acceptable)
46 ACK from the destination. If this number is reached, a shorter timer
48 static int nf_ct_tcp_max_retrans __read_mostly
= 3;
50 /* FIXME: Examine ipfilter's timeouts and conntrack transitions more
51 closely. They're more complex. --RR */
53 static const char *const tcp_conntrack_names
[] = {
67 #define MINS * 60 SECS
68 #define HOURS * 60 MINS
69 #define DAYS * 24 HOURS
71 static unsigned int tcp_timeouts
[TCP_CONNTRACK_TIMEOUT_MAX
] __read_mostly
= {
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
,
80 [TCP_CONNTRACK_SYN_SENT2
] = 2 MINS
,
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
,
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
97 #define sS2 TCP_CONNTRACK_SYN_SENT2
98 #define sIV TCP_CONNTRACK_MAX
99 #define sIG TCP_CONNTRACK_IGNORE
101 /* What TCP flags are set from RST/SYN/FIN/ACK. */
112 * The TCP state transition table needs a few words...
114 * We are the man in the middle. All the packets go through us
115 * but might get lost in transit to the destination.
116 * It is assumed that the destinations can't receive segments
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.
123 * The meaning of the states are:
125 * NONE: initial state
126 * SYN_SENT: SYN-only packet seen
127 * SYN_SENT2: SYN-only packet seen from reply dir, simultaneous open
128 * SYN_RECV: SYN-ACK packet seen
129 * ESTABLISHED: ACK packet seen
130 * FIN_WAIT: FIN packet seen
131 * CLOSE_WAIT: ACK seen (after FIN)
132 * LAST_ACK: FIN seen (after FIN)
133 * TIME_WAIT: last ACK seen
134 * CLOSE: closed connection (RST)
136 * Packets marked as IGNORED (sIG):
137 * if they may be either invalid or valid
138 * and the receiver may send back a connection
139 * closing RST or a SYN/ACK.
141 * Packets marked as INVALID (sIV):
142 * if we regard them as truly invalid packets
144 static const u8 tcp_conntracks
[2][6][TCP_CONNTRACK_MAX
] = {
147 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
148 /*syn*/ { sSS
, sSS
, sIG
, sIG
, sIG
, sIG
, sIG
, sSS
, sSS
, sS2
},
150 * sNO -> sSS Initialize a new connection
151 * sSS -> sSS Retransmitted SYN
152 * sS2 -> sS2 Late retransmitted SYN
154 * sES -> sIG Error: SYNs in window outside the SYN_SENT state
155 * are errors. Receiver will reply with RST
156 * and close the connection.
157 * Or we are not in sync and hold a dead connection.
161 * sTW -> sSS Reopened connection (RFC 1122).
164 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
165 /*synack*/ { sIV
, sIV
, sSR
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sSR
},
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
170 * sSR -> sSR Late retransmitted SYN/ACK in simultaneous open
171 * sES -> sIV Invalid SYN/ACK packets sent by the client
178 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
179 /*fin*/ { sIV
, sIV
, sFW
, sFW
, sLA
, sLA
, sLA
, sTW
, sCL
, sIV
},
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.
185 * sSR -> sFW Close started.
187 * sFW -> sLA FIN seen in both directions, waiting for
189 * Migth be a retransmitted FIN as well...
191 * sLA -> sLA Retransmitted FIN. Remain in the same state.
195 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
196 /*ack*/ { sES
, sIV
, sES
, sES
, sCW
, sCW
, sTW
, sTW
, sCL
, sIV
},
198 * sNO -> sES Assumed.
199 * sSS -> sIV ACK is invalid: we haven't seen a SYN/ACK yet.
201 * sSR -> sES Established state is reached.
203 * sFW -> sCW Normal close request answered by ACK.
205 * sLA -> sTW Last ACK detected (RFC5961 challenged)
206 * sTW -> sTW Retransmitted last ACK. Remain in the same state.
209 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
210 /*rst*/ { sIV
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
},
211 /*none*/ { sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
}
215 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
216 /*syn*/ { sIV
, sS2
, sIV
, sIV
, sIV
, sIV
, sIV
, sSS
, sIV
, sS2
},
218 * sNO -> sIV Never reached.
219 * sSS -> sS2 Simultaneous open
220 * sS2 -> sS2 Retransmitted simultaneous SYN
221 * sSR -> sIV Invalid SYN packets sent by the server
226 * sTW -> sSS Reopened connection, but server may have switched role
229 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
230 /*synack*/ { sIV
, sSR
, sIG
, sIG
, sIG
, sIG
, sIG
, sIG
, sIG
, sSR
},
232 * sSS -> sSR Standard open.
233 * sS2 -> sSR Simultaneous open
234 * sSR -> sIG Retransmitted SYN/ACK, ignore it.
235 * sES -> sIG Late retransmitted SYN/ACK?
236 * sFW -> sIG Might be SYN/ACK answering ignored SYN
242 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
243 /*fin*/ { sIV
, sIV
, sFW
, sFW
, sLA
, sLA
, sLA
, sTW
, sCL
, sIV
},
245 * sSS -> sIV Server might not send FIN in this state.
247 * sSR -> sFW Close started.
249 * sFW -> sLA FIN seen in both directions.
251 * sLA -> sLA Retransmitted FIN.
255 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
256 /*ack*/ { sIV
, sIG
, sSR
, sES
, sCW
, sCW
, sTW
, sTW
, sCL
, sIG
},
258 * sSS -> sIG Might be a half-open connection.
260 * sSR -> sSR Might answer late resent SYN.
262 * sFW -> sCW Normal close request answered by ACK.
264 * sLA -> sTW Last ACK detected (RFC5961 challenged)
265 * sTW -> sTW Retransmitted last ACK.
268 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
269 /*rst*/ { sIV
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
, sCL
},
270 /*none*/ { sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
, sIV
}
274 static inline struct nf_tcp_net
*tcp_pernet(struct net
*net
)
276 return &net
->ct
.nf_ct_proto
.tcp
;
279 static bool tcp_pkt_to_tuple(const struct sk_buff
*skb
, unsigned int dataoff
,
280 struct net
*net
, struct nf_conntrack_tuple
*tuple
)
282 const struct tcphdr
*hp
;
285 /* Actually only need first 4 bytes to get ports. */
286 hp
= skb_header_pointer(skb
, dataoff
, 4, &_hdr
);
290 tuple
->src
.u
.tcp
.port
= hp
->source
;
291 tuple
->dst
.u
.tcp
.port
= hp
->dest
;
296 static bool tcp_invert_tuple(struct nf_conntrack_tuple
*tuple
,
297 const struct nf_conntrack_tuple
*orig
)
299 tuple
->src
.u
.tcp
.port
= orig
->dst
.u
.tcp
.port
;
300 tuple
->dst
.u
.tcp
.port
= orig
->src
.u
.tcp
.port
;
304 #ifdef CONFIG_NF_CONNTRACK_PROCFS
305 /* Print out the private part of the conntrack. */
306 static void tcp_print_conntrack(struct seq_file
*s
, struct nf_conn
*ct
)
308 seq_printf(s
, "%s ", tcp_conntrack_names
[ct
->proto
.tcp
.state
]);
312 static unsigned int get_conntrack_index(const struct tcphdr
*tcph
)
314 if (tcph
->rst
) return TCP_RST_SET
;
315 else if (tcph
->syn
) return (tcph
->ack
? TCP_SYNACK_SET
: TCP_SYN_SET
);
316 else if (tcph
->fin
) return TCP_FIN_SET
;
317 else if (tcph
->ack
) return TCP_ACK_SET
;
318 else return TCP_NONE_SET
;
321 /* TCP connection tracking based on 'Real Stateful TCP Packet Filtering
322 in IP Filter' by Guido van Rooij.
324 http://www.sane.nl/events/sane2000/papers.html
325 http://www.darkart.com/mirrors/www.obfuscation.org/ipf/
327 The boundaries and the conditions are changed according to RFC793:
328 the packet must intersect the window (i.e. segments may be
329 after the right or before the left edge) and thus receivers may ACK
330 segments after the right edge of the window.
332 td_maxend = max(sack + max(win,1)) seen in reply packets
333 td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets
334 td_maxwin += seq + len - sender.td_maxend
335 if seq + len > sender.td_maxend
336 td_end = max(seq + len) seen in sent packets
338 I. Upper bound for valid data: seq <= sender.td_maxend
339 II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin
340 III. Upper bound for valid (s)ack: sack <= receiver.td_end
341 IV. Lower bound for valid (s)ack: sack >= receiver.td_end - MAXACKWINDOW
343 where sack is the highest right edge of sack block found in the packet
344 or ack in the case of packet without SACK option.
346 The upper bound limit for a valid (s)ack is not ignored -
347 we doesn't have to deal with fragments.
350 static inline __u32
segment_seq_plus_len(__u32 seq
,
352 unsigned int dataoff
,
353 const struct tcphdr
*tcph
)
355 /* XXX Should I use payload length field in IP/IPv6 header ?
357 return (seq
+ len
- dataoff
- tcph
->doff
*4
358 + (tcph
->syn
? 1 : 0) + (tcph
->fin
? 1 : 0));
361 /* Fixme: what about big packets? */
362 #define MAXACKWINCONST 66000
363 #define MAXACKWINDOW(sender) \
364 ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \
368 * Simplified tcp_parse_options routine from tcp_input.c
370 static void tcp_options(const struct sk_buff
*skb
,
371 unsigned int dataoff
,
372 const struct tcphdr
*tcph
,
373 struct ip_ct_tcp_state
*state
)
375 unsigned char buff
[(15 * 4) - sizeof(struct tcphdr
)];
376 const unsigned char *ptr
;
377 int length
= (tcph
->doff
*4) - sizeof(struct tcphdr
);
382 ptr
= skb_header_pointer(skb
, dataoff
+ sizeof(struct tcphdr
),
396 case TCPOPT_NOP
: /* Ref: RFC 793 section 3.1 */
403 if (opsize
< 2) /* "silly options" */
406 return; /* don't parse partial options */
408 if (opcode
== TCPOPT_SACK_PERM
409 && opsize
== TCPOLEN_SACK_PERM
)
410 state
->flags
|= IP_CT_TCP_FLAG_SACK_PERM
;
411 else if (opcode
== TCPOPT_WINDOW
412 && opsize
== TCPOLEN_WINDOW
) {
413 state
->td_scale
= *(u_int8_t
*)ptr
;
415 if (state
->td_scale
> TCP_MAX_WSCALE
)
416 state
->td_scale
= TCP_MAX_WSCALE
;
419 IP_CT_TCP_FLAG_WINDOW_SCALE
;
427 static void tcp_sack(const struct sk_buff
*skb
, unsigned int dataoff
,
428 const struct tcphdr
*tcph
, __u32
*sack
)
430 unsigned char buff
[(15 * 4) - sizeof(struct tcphdr
)];
431 const unsigned char *ptr
;
432 int length
= (tcph
->doff
*4) - sizeof(struct tcphdr
);
438 ptr
= skb_header_pointer(skb
, dataoff
+ sizeof(struct tcphdr
),
442 /* Fast path for timestamp-only option */
443 if (length
== TCPOLEN_TSTAMP_ALIGNED
444 && *(__be32
*)ptr
== htonl((TCPOPT_NOP
<< 24)
446 | (TCPOPT_TIMESTAMP
<< 8)
447 | TCPOLEN_TIMESTAMP
))
457 case TCPOPT_NOP
: /* Ref: RFC 793 section 3.1 */
464 if (opsize
< 2) /* "silly options" */
467 return; /* don't parse partial options */
469 if (opcode
== TCPOPT_SACK
470 && opsize
>= (TCPOLEN_SACK_BASE
471 + TCPOLEN_SACK_PERBLOCK
)
472 && !((opsize
- TCPOLEN_SACK_BASE
)
473 % TCPOLEN_SACK_PERBLOCK
)) {
475 i
< (opsize
- TCPOLEN_SACK_BASE
);
476 i
+= TCPOLEN_SACK_PERBLOCK
) {
477 tmp
= get_unaligned_be32((__be32
*)(ptr
+i
)+1);
479 if (after(tmp
, *sack
))
490 static bool tcp_in_window(const struct nf_conn
*ct
,
491 struct ip_ct_tcp
*state
,
492 enum ip_conntrack_dir dir
,
494 const struct sk_buff
*skb
,
495 unsigned int dataoff
,
496 const struct tcphdr
*tcph
)
498 struct net
*net
= nf_ct_net(ct
);
499 struct nf_tcp_net
*tn
= tcp_pernet(net
);
500 struct ip_ct_tcp_state
*sender
= &state
->seen
[dir
];
501 struct ip_ct_tcp_state
*receiver
= &state
->seen
[!dir
];
502 const struct nf_conntrack_tuple
*tuple
= &ct
->tuplehash
[dir
].tuple
;
503 __u32 seq
, ack
, sack
, end
, win
, swin
;
505 bool res
, in_recv_win
;
508 * Get the required data from the packet.
510 seq
= ntohl(tcph
->seq
);
511 ack
= sack
= ntohl(tcph
->ack_seq
);
512 win
= ntohs(tcph
->window
);
513 end
= segment_seq_plus_len(seq
, skb
->len
, dataoff
, tcph
);
515 if (receiver
->flags
& IP_CT_TCP_FLAG_SACK_PERM
)
516 tcp_sack(skb
, dataoff
, tcph
, &sack
);
518 /* Take into account NAT sequence number mangling */
519 receiver_offset
= nf_ct_seq_offset(ct
, !dir
, ack
- 1);
520 ack
-= receiver_offset
;
521 sack
-= receiver_offset
;
523 pr_debug("tcp_in_window: START\n");
524 pr_debug("tcp_in_window: ");
525 nf_ct_dump_tuple(tuple
);
526 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
527 seq
, ack
, receiver_offset
, sack
, receiver_offset
, win
, end
);
528 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
529 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
530 sender
->td_end
, sender
->td_maxend
, sender
->td_maxwin
,
532 receiver
->td_end
, receiver
->td_maxend
, receiver
->td_maxwin
,
535 if (sender
->td_maxwin
== 0) {
537 * Initialize sender data.
541 * SYN-ACK in reply to a SYN
542 * or SYN from reply direction in simultaneous open.
545 sender
->td_maxend
= end
;
546 sender
->td_maxwin
= (win
== 0 ? 1 : win
);
548 tcp_options(skb
, dataoff
, tcph
, sender
);
551 * Both sides must send the Window Scale option
552 * to enable window scaling in either direction.
554 if (!(sender
->flags
& IP_CT_TCP_FLAG_WINDOW_SCALE
555 && receiver
->flags
& IP_CT_TCP_FLAG_WINDOW_SCALE
))
557 receiver
->td_scale
= 0;
559 /* Simultaneous open */
563 * We are in the middle of a connection,
564 * its history is lost for us.
565 * Let's try to use the data from the packet.
567 sender
->td_end
= end
;
568 swin
= win
<< sender
->td_scale
;
569 sender
->td_maxwin
= (swin
== 0 ? 1 : swin
);
570 sender
->td_maxend
= end
+ sender
->td_maxwin
;
572 * We haven't seen traffic in the other direction yet
573 * but we have to tweak window tracking to pass III
574 * and IV until that happens.
576 if (receiver
->td_maxwin
== 0)
577 receiver
->td_end
= receiver
->td_maxend
= sack
;
579 } else if (((state
->state
== TCP_CONNTRACK_SYN_SENT
580 && dir
== IP_CT_DIR_ORIGINAL
)
581 || (state
->state
== TCP_CONNTRACK_SYN_RECV
582 && dir
== IP_CT_DIR_REPLY
))
583 && after(end
, sender
->td_end
)) {
585 * RFC 793: "if a TCP is reinitialized ... then it need
586 * not wait at all; it must only be sure to use sequence
587 * numbers larger than those recently used."
590 sender
->td_maxend
= end
;
591 sender
->td_maxwin
= (win
== 0 ? 1 : win
);
593 tcp_options(skb
, dataoff
, tcph
, sender
);
598 * If there is no ACK, just pretend it was set and OK.
600 ack
= sack
= receiver
->td_end
;
601 } else if (((tcp_flag_word(tcph
) & (TCP_FLAG_ACK
|TCP_FLAG_RST
)) ==
602 (TCP_FLAG_ACK
|TCP_FLAG_RST
))
605 * Broken TCP stacks, that set ACK in RST packets as well
606 * with zero ack value.
608 ack
= sack
= receiver
->td_end
;
611 if (tcph
->rst
&& seq
== 0 && state
->state
== TCP_CONNTRACK_SYN_SENT
)
613 * RST sent answering SYN.
615 seq
= end
= sender
->td_end
;
617 pr_debug("tcp_in_window: ");
618 nf_ct_dump_tuple(tuple
);
619 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
620 seq
, ack
, receiver_offset
, sack
, receiver_offset
, win
, end
);
621 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
622 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
623 sender
->td_end
, sender
->td_maxend
, sender
->td_maxwin
,
625 receiver
->td_end
, receiver
->td_maxend
, receiver
->td_maxwin
,
628 /* Is the ending sequence in the receive window (if available)? */
629 in_recv_win
= !receiver
->td_maxwin
||
630 after(end
, sender
->td_end
- receiver
->td_maxwin
- 1);
632 pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
633 before(seq
, sender
->td_maxend
+ 1),
634 (in_recv_win
? 1 : 0),
635 before(sack
, receiver
->td_end
+ 1),
636 after(sack
, receiver
->td_end
- MAXACKWINDOW(sender
) - 1));
638 if (before(seq
, sender
->td_maxend
+ 1) &&
640 before(sack
, receiver
->td_end
+ 1) &&
641 after(sack
, receiver
->td_end
- MAXACKWINDOW(sender
) - 1)) {
643 * Take into account window scaling (RFC 1323).
646 win
<<= sender
->td_scale
;
649 * Update sender data.
651 swin
= win
+ (sack
- ack
);
652 if (sender
->td_maxwin
< swin
)
653 sender
->td_maxwin
= swin
;
654 if (after(end
, sender
->td_end
)) {
655 sender
->td_end
= end
;
656 sender
->flags
|= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED
;
659 if (!(sender
->flags
& IP_CT_TCP_FLAG_MAXACK_SET
)) {
660 sender
->td_maxack
= ack
;
661 sender
->flags
|= IP_CT_TCP_FLAG_MAXACK_SET
;
662 } else if (after(ack
, sender
->td_maxack
))
663 sender
->td_maxack
= ack
;
667 * Update receiver data.
669 if (receiver
->td_maxwin
!= 0 && after(end
, sender
->td_maxend
))
670 receiver
->td_maxwin
+= end
- sender
->td_maxend
;
671 if (after(sack
+ win
, receiver
->td_maxend
- 1)) {
672 receiver
->td_maxend
= sack
+ win
;
674 receiver
->td_maxend
++;
676 if (ack
== receiver
->td_end
)
677 receiver
->flags
&= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED
;
680 * Check retransmissions.
682 if (index
== TCP_ACK_SET
) {
683 if (state
->last_dir
== dir
684 && state
->last_seq
== seq
685 && state
->last_ack
== ack
686 && state
->last_end
== end
687 && state
->last_win
== win
)
690 state
->last_dir
= dir
;
691 state
->last_seq
= seq
;
692 state
->last_ack
= ack
;
693 state
->last_end
= end
;
694 state
->last_win
= win
;
701 if (sender
->flags
& IP_CT_TCP_FLAG_BE_LIBERAL
||
705 nf_ct_l4proto_log_invalid(skb
, ct
,
707 before(seq
, sender
->td_maxend
+ 1) ?
709 before(sack
, receiver
->td_end
+ 1) ?
710 after(sack
, receiver
->td_end
- MAXACKWINDOW(sender
) - 1) ? "BUG"
711 : "ACK is under the lower bound (possible overly delayed ACK)"
712 : "ACK is over the upper bound (ACKed data not seen yet)"
713 : "SEQ is under the lower bound (already ACKed data retransmitted)"
714 : "SEQ is over the upper bound (over the window of the receiver)");
718 pr_debug("tcp_in_window: res=%u sender end=%u maxend=%u maxwin=%u "
719 "receiver end=%u maxend=%u maxwin=%u\n",
720 res
, sender
->td_end
, sender
->td_maxend
, sender
->td_maxwin
,
721 receiver
->td_end
, receiver
->td_maxend
, receiver
->td_maxwin
);
726 /* table of valid flag combinations - PUSH, ECE and CWR are always valid */
727 static const u8 tcp_valid_flags
[(TCPHDR_FIN
|TCPHDR_SYN
|TCPHDR_RST
|TCPHDR_ACK
|
731 [TCPHDR_SYN
|TCPHDR_URG
] = 1,
732 [TCPHDR_SYN
|TCPHDR_ACK
] = 1,
734 [TCPHDR_RST
|TCPHDR_ACK
] = 1,
735 [TCPHDR_FIN
|TCPHDR_ACK
] = 1,
736 [TCPHDR_FIN
|TCPHDR_ACK
|TCPHDR_URG
] = 1,
738 [TCPHDR_ACK
|TCPHDR_URG
] = 1,
741 static void tcp_error_log(const struct sk_buff
*skb
, struct net
*net
,
742 u8 pf
, const char *msg
)
744 nf_l4proto_log_invalid(skb
, net
, pf
, IPPROTO_TCP
, "%s", msg
);
747 /* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
748 static int tcp_error(struct net
*net
, struct nf_conn
*tmpl
,
750 unsigned int dataoff
,
752 unsigned int hooknum
)
754 const struct tcphdr
*th
;
756 unsigned int tcplen
= skb
->len
- dataoff
;
759 /* Smaller that minimal TCP header? */
760 th
= skb_header_pointer(skb
, dataoff
, sizeof(_tcph
), &_tcph
);
762 tcp_error_log(skb
, net
, pf
, "short packet");
766 /* Not whole TCP header or malformed packet */
767 if (th
->doff
*4 < sizeof(struct tcphdr
) || tcplen
< th
->doff
*4) {
768 tcp_error_log(skb
, net
, pf
, "truncated packet");
772 /* Checksum invalid? Ignore.
773 * We skip checking packets on the outgoing path
774 * because the checksum is assumed to be correct.
776 /* FIXME: Source route IP option packets --RR */
777 if (net
->ct
.sysctl_checksum
&& hooknum
== NF_INET_PRE_ROUTING
&&
778 nf_checksum(skb
, hooknum
, dataoff
, IPPROTO_TCP
, pf
)) {
779 tcp_error_log(skb
, net
, pf
, "bad checksum");
783 /* Check TCP flags. */
784 tcpflags
= (tcp_flag_byte(th
) & ~(TCPHDR_ECE
|TCPHDR_CWR
|TCPHDR_PSH
));
785 if (!tcp_valid_flags
[tcpflags
]) {
786 tcp_error_log(skb
, net
, pf
, "invalid tcp flag combination");
793 static unsigned int *tcp_get_timeouts(struct net
*net
)
795 return tcp_pernet(net
)->timeouts
;
798 /* Returns verdict for packet, or -1 for invalid. */
799 static int tcp_packet(struct nf_conn
*ct
,
800 const struct sk_buff
*skb
,
801 unsigned int dataoff
,
802 enum ip_conntrack_info ctinfo
,
803 unsigned int *timeouts
)
805 struct net
*net
= nf_ct_net(ct
);
806 struct nf_tcp_net
*tn
= tcp_pernet(net
);
807 struct nf_conntrack_tuple
*tuple
;
808 enum tcp_conntrack new_state
, old_state
;
809 enum ip_conntrack_dir dir
;
810 const struct tcphdr
*th
;
812 unsigned long timeout
;
815 th
= skb_header_pointer(skb
, dataoff
, sizeof(_tcph
), &_tcph
);
818 spin_lock_bh(&ct
->lock
);
819 old_state
= ct
->proto
.tcp
.state
;
820 dir
= CTINFO2DIR(ctinfo
);
821 index
= get_conntrack_index(th
);
822 new_state
= tcp_conntracks
[dir
][index
][old_state
];
823 tuple
= &ct
->tuplehash
[dir
].tuple
;
826 case TCP_CONNTRACK_SYN_SENT
:
827 if (old_state
< TCP_CONNTRACK_TIME_WAIT
)
829 /* RFC 1122: "When a connection is closed actively,
830 * it MUST linger in TIME-WAIT state for a time 2xMSL
831 * (Maximum Segment Lifetime). However, it MAY accept
832 * a new SYN from the remote TCP to reopen the connection
833 * directly from TIME-WAIT state, if..."
834 * We ignore the conditions because we are in the
835 * TIME-WAIT state anyway.
837 * Handle aborted connections: we and the server
838 * think there is an existing connection but the client
839 * aborts it and starts a new one.
841 if (((ct
->proto
.tcp
.seen
[dir
].flags
842 | ct
->proto
.tcp
.seen
[!dir
].flags
)
843 & IP_CT_TCP_FLAG_CLOSE_INIT
)
844 || (ct
->proto
.tcp
.last_dir
== dir
845 && ct
->proto
.tcp
.last_index
== TCP_RST_SET
)) {
846 /* Attempt to reopen a closed/aborted connection.
847 * Delete this connection and look up again. */
848 spin_unlock_bh(&ct
->lock
);
850 /* Only repeat if we can actually remove the timer.
851 * Destruction may already be in progress in process
852 * context and we must give it a chance to terminate.
859 case TCP_CONNTRACK_IGNORE
:
862 * Our connection entry may be out of sync, so ignore
863 * packets which may signal the real connection between
864 * the client and the server.
867 * b) SYN/ACK in REPLY
868 * c) ACK in reply direction after initial SYN in original.
870 * If the ignored packet is invalid, the receiver will send
871 * a RST we'll catch below.
873 if (index
== TCP_SYNACK_SET
874 && ct
->proto
.tcp
.last_index
== TCP_SYN_SET
875 && ct
->proto
.tcp
.last_dir
!= dir
876 && ntohl(th
->ack_seq
) == ct
->proto
.tcp
.last_end
) {
877 /* b) This SYN/ACK acknowledges a SYN that we earlier
878 * ignored as invalid. This means that the client and
879 * the server are both in sync, while the firewall is
880 * not. We get in sync from the previously annotated
883 old_state
= TCP_CONNTRACK_SYN_SENT
;
884 new_state
= TCP_CONNTRACK_SYN_RECV
;
885 ct
->proto
.tcp
.seen
[ct
->proto
.tcp
.last_dir
].td_end
=
886 ct
->proto
.tcp
.last_end
;
887 ct
->proto
.tcp
.seen
[ct
->proto
.tcp
.last_dir
].td_maxend
=
888 ct
->proto
.tcp
.last_end
;
889 ct
->proto
.tcp
.seen
[ct
->proto
.tcp
.last_dir
].td_maxwin
=
890 ct
->proto
.tcp
.last_win
== 0 ?
891 1 : ct
->proto
.tcp
.last_win
;
892 ct
->proto
.tcp
.seen
[ct
->proto
.tcp
.last_dir
].td_scale
=
893 ct
->proto
.tcp
.last_wscale
;
894 ct
->proto
.tcp
.last_flags
&= ~IP_CT_EXP_CHALLENGE_ACK
;
895 ct
->proto
.tcp
.seen
[ct
->proto
.tcp
.last_dir
].flags
=
896 ct
->proto
.tcp
.last_flags
;
897 memset(&ct
->proto
.tcp
.seen
[dir
], 0,
898 sizeof(struct ip_ct_tcp_state
));
901 ct
->proto
.tcp
.last_index
= index
;
902 ct
->proto
.tcp
.last_dir
= dir
;
903 ct
->proto
.tcp
.last_seq
= ntohl(th
->seq
);
904 ct
->proto
.tcp
.last_end
=
905 segment_seq_plus_len(ntohl(th
->seq
), skb
->len
, dataoff
, th
);
906 ct
->proto
.tcp
.last_win
= ntohs(th
->window
);
908 /* a) This is a SYN in ORIGINAL. The client and the server
909 * may be in sync but we are not. In that case, we annotate
910 * the TCP options and let the packet go through. If it is a
911 * valid SYN packet, the server will reply with a SYN/ACK, and
912 * then we'll get in sync. Otherwise, the server potentially
913 * responds with a challenge ACK if implementing RFC5961.
915 if (index
== TCP_SYN_SET
&& dir
== IP_CT_DIR_ORIGINAL
) {
916 struct ip_ct_tcp_state seen
= {};
918 ct
->proto
.tcp
.last_flags
=
919 ct
->proto
.tcp
.last_wscale
= 0;
920 tcp_options(skb
, dataoff
, th
, &seen
);
921 if (seen
.flags
& IP_CT_TCP_FLAG_WINDOW_SCALE
) {
922 ct
->proto
.tcp
.last_flags
|=
923 IP_CT_TCP_FLAG_WINDOW_SCALE
;
924 ct
->proto
.tcp
.last_wscale
= seen
.td_scale
;
926 if (seen
.flags
& IP_CT_TCP_FLAG_SACK_PERM
) {
927 ct
->proto
.tcp
.last_flags
|=
928 IP_CT_TCP_FLAG_SACK_PERM
;
930 /* Mark the potential for RFC5961 challenge ACK,
931 * this pose a special problem for LAST_ACK state
932 * as ACK is intrepretated as ACKing last FIN.
934 if (old_state
== TCP_CONNTRACK_LAST_ACK
)
935 ct
->proto
.tcp
.last_flags
|=
936 IP_CT_EXP_CHALLENGE_ACK
;
938 spin_unlock_bh(&ct
->lock
);
939 nf_ct_l4proto_log_invalid(skb
, ct
, "invalid packet ignored in "
940 "state %s ", tcp_conntrack_names
[old_state
]);
942 case TCP_CONNTRACK_MAX
:
943 /* Special case for SYN proxy: when the SYN to the server or
944 * the SYN/ACK from the server is lost, the client may transmit
945 * a keep-alive packet while in SYN_SENT state. This needs to
946 * be associated with the original conntrack entry in order to
947 * generate a new SYN with the correct sequence number.
949 if (nfct_synproxy(ct
) && old_state
== TCP_CONNTRACK_SYN_SENT
&&
950 index
== TCP_ACK_SET
&& dir
== IP_CT_DIR_ORIGINAL
&&
951 ct
->proto
.tcp
.last_dir
== IP_CT_DIR_ORIGINAL
&&
952 ct
->proto
.tcp
.seen
[dir
].td_end
- 1 == ntohl(th
->seq
)) {
953 pr_debug("nf_ct_tcp: SYN proxy client keep alive\n");
954 spin_unlock_bh(&ct
->lock
);
959 pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
960 dir
, get_conntrack_index(th
), old_state
);
961 spin_unlock_bh(&ct
->lock
);
962 nf_ct_l4proto_log_invalid(skb
, ct
, "invalid state");
964 case TCP_CONNTRACK_TIME_WAIT
:
965 /* RFC5961 compliance cause stack to send "challenge-ACK"
966 * e.g. in response to spurious SYNs. Conntrack MUST
967 * not believe this ACK is acking last FIN.
969 if (old_state
== TCP_CONNTRACK_LAST_ACK
&&
970 index
== TCP_ACK_SET
&&
971 ct
->proto
.tcp
.last_dir
!= dir
&&
972 ct
->proto
.tcp
.last_index
== TCP_SYN_SET
&&
973 (ct
->proto
.tcp
.last_flags
& IP_CT_EXP_CHALLENGE_ACK
)) {
974 /* Detected RFC5961 challenge ACK */
975 ct
->proto
.tcp
.last_flags
&= ~IP_CT_EXP_CHALLENGE_ACK
;
976 spin_unlock_bh(&ct
->lock
);
977 nf_ct_l4proto_log_invalid(skb
, ct
, "challenge-ack ignored");
978 return NF_ACCEPT
; /* Don't change state */
981 case TCP_CONNTRACK_CLOSE
:
982 if (index
== TCP_RST_SET
983 && (ct
->proto
.tcp
.seen
[!dir
].flags
& IP_CT_TCP_FLAG_MAXACK_SET
)
984 && before(ntohl(th
->seq
), ct
->proto
.tcp
.seen
[!dir
].td_maxack
)) {
986 spin_unlock_bh(&ct
->lock
);
987 nf_ct_l4proto_log_invalid(skb
, ct
, "invalid rst");
990 if (index
== TCP_RST_SET
991 && ((test_bit(IPS_SEEN_REPLY_BIT
, &ct
->status
)
992 && ct
->proto
.tcp
.last_index
== TCP_SYN_SET
)
993 || (!test_bit(IPS_ASSURED_BIT
, &ct
->status
)
994 && ct
->proto
.tcp
.last_index
== TCP_ACK_SET
))
995 && ntohl(th
->ack_seq
) == ct
->proto
.tcp
.last_end
) {
996 /* RST sent to invalid SYN or ACK we had let through
997 * at a) and c) above:
999 * a) SYN was in window then
1000 * c) we hold a half-open connection.
1002 * Delete our connection entry.
1003 * We skip window checking, because packet might ACK
1004 * segments we ignored. */
1007 /* Just fall through */
1009 /* Keep compilers happy. */
1013 if (!tcp_in_window(ct
, &ct
->proto
.tcp
, dir
, index
,
1014 skb
, dataoff
, th
)) {
1015 spin_unlock_bh(&ct
->lock
);
1019 /* From now on we have got in-window packets */
1020 ct
->proto
.tcp
.last_index
= index
;
1021 ct
->proto
.tcp
.last_dir
= dir
;
1023 pr_debug("tcp_conntracks: ");
1024 nf_ct_dump_tuple(tuple
);
1025 pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
1026 (th
->syn
? 1 : 0), (th
->ack
? 1 : 0),
1027 (th
->fin
? 1 : 0), (th
->rst
? 1 : 0),
1028 old_state
, new_state
);
1030 ct
->proto
.tcp
.state
= new_state
;
1031 if (old_state
!= new_state
1032 && new_state
== TCP_CONNTRACK_FIN_WAIT
)
1033 ct
->proto
.tcp
.seen
[dir
].flags
|= IP_CT_TCP_FLAG_CLOSE_INIT
;
1035 if (ct
->proto
.tcp
.retrans
>= tn
->tcp_max_retrans
&&
1036 timeouts
[new_state
] > timeouts
[TCP_CONNTRACK_RETRANS
])
1037 timeout
= timeouts
[TCP_CONNTRACK_RETRANS
];
1038 else if ((ct
->proto
.tcp
.seen
[0].flags
| ct
->proto
.tcp
.seen
[1].flags
) &
1039 IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED
&&
1040 timeouts
[new_state
] > timeouts
[TCP_CONNTRACK_UNACK
])
1041 timeout
= timeouts
[TCP_CONNTRACK_UNACK
];
1042 else if (ct
->proto
.tcp
.last_win
== 0 &&
1043 timeouts
[new_state
] > timeouts
[TCP_CONNTRACK_RETRANS
])
1044 timeout
= timeouts
[TCP_CONNTRACK_RETRANS
];
1046 timeout
= timeouts
[new_state
];
1047 spin_unlock_bh(&ct
->lock
);
1049 if (new_state
!= old_state
)
1050 nf_conntrack_event_cache(IPCT_PROTOINFO
, ct
);
1052 if (!test_bit(IPS_SEEN_REPLY_BIT
, &ct
->status
)) {
1053 /* If only reply is a RST, we can consider ourselves not to
1054 have an established connection: this is a fairly common
1055 problem case, so we can delete the conntrack
1056 immediately. --RR */
1058 nf_ct_kill_acct(ct
, ctinfo
, skb
);
1061 /* ESTABLISHED without SEEN_REPLY, i.e. mid-connection
1062 * pickup with loose=1. Avoid large ESTABLISHED timeout.
1064 if (new_state
== TCP_CONNTRACK_ESTABLISHED
&&
1065 timeout
> timeouts
[TCP_CONNTRACK_UNACK
])
1066 timeout
= timeouts
[TCP_CONNTRACK_UNACK
];
1067 } else if (!test_bit(IPS_ASSURED_BIT
, &ct
->status
)
1068 && (old_state
== TCP_CONNTRACK_SYN_RECV
1069 || old_state
== TCP_CONNTRACK_ESTABLISHED
)
1070 && new_state
== TCP_CONNTRACK_ESTABLISHED
) {
1071 /* Set ASSURED if we see see valid ack in ESTABLISHED
1072 after SYN_RECV or a valid answer for a picked up
1074 set_bit(IPS_ASSURED_BIT
, &ct
->status
);
1075 nf_conntrack_event_cache(IPCT_ASSURED
, ct
);
1077 nf_ct_refresh_acct(ct
, ctinfo
, skb
, timeout
);
1082 /* Called when a new connection for this protocol found. */
1083 static bool tcp_new(struct nf_conn
*ct
, const struct sk_buff
*skb
,
1084 unsigned int dataoff
, unsigned int *timeouts
)
1086 enum tcp_conntrack new_state
;
1087 const struct tcphdr
*th
;
1088 struct tcphdr _tcph
;
1089 struct net
*net
= nf_ct_net(ct
);
1090 struct nf_tcp_net
*tn
= tcp_pernet(net
);
1091 const struct ip_ct_tcp_state
*sender
= &ct
->proto
.tcp
.seen
[0];
1092 const struct ip_ct_tcp_state
*receiver
= &ct
->proto
.tcp
.seen
[1];
1094 th
= skb_header_pointer(skb
, dataoff
, sizeof(_tcph
), &_tcph
);
1097 /* Don't need lock here: this conntrack not in circulation yet */
1098 new_state
= tcp_conntracks
[0][get_conntrack_index(th
)][TCP_CONNTRACK_NONE
];
1100 /* Invalid: delete conntrack */
1101 if (new_state
>= TCP_CONNTRACK_MAX
) {
1102 pr_debug("nf_ct_tcp: invalid new deleting.\n");
1106 if (new_state
== TCP_CONNTRACK_SYN_SENT
) {
1107 memset(&ct
->proto
.tcp
, 0, sizeof(ct
->proto
.tcp
));
1109 ct
->proto
.tcp
.seen
[0].td_end
=
1110 segment_seq_plus_len(ntohl(th
->seq
), skb
->len
,
1112 ct
->proto
.tcp
.seen
[0].td_maxwin
= ntohs(th
->window
);
1113 if (ct
->proto
.tcp
.seen
[0].td_maxwin
== 0)
1114 ct
->proto
.tcp
.seen
[0].td_maxwin
= 1;
1115 ct
->proto
.tcp
.seen
[0].td_maxend
=
1116 ct
->proto
.tcp
.seen
[0].td_end
;
1118 tcp_options(skb
, dataoff
, th
, &ct
->proto
.tcp
.seen
[0]);
1119 } else if (tn
->tcp_loose
== 0) {
1120 /* Don't try to pick up connections. */
1123 memset(&ct
->proto
.tcp
, 0, sizeof(ct
->proto
.tcp
));
1125 * We are in the middle of a connection,
1126 * its history is lost for us.
1127 * Let's try to use the data from the packet.
1129 ct
->proto
.tcp
.seen
[0].td_end
=
1130 segment_seq_plus_len(ntohl(th
->seq
), skb
->len
,
1132 ct
->proto
.tcp
.seen
[0].td_maxwin
= ntohs(th
->window
);
1133 if (ct
->proto
.tcp
.seen
[0].td_maxwin
== 0)
1134 ct
->proto
.tcp
.seen
[0].td_maxwin
= 1;
1135 ct
->proto
.tcp
.seen
[0].td_maxend
=
1136 ct
->proto
.tcp
.seen
[0].td_end
+
1137 ct
->proto
.tcp
.seen
[0].td_maxwin
;
1139 /* We assume SACK and liberal window checking to handle
1141 ct
->proto
.tcp
.seen
[0].flags
=
1142 ct
->proto
.tcp
.seen
[1].flags
= IP_CT_TCP_FLAG_SACK_PERM
|
1143 IP_CT_TCP_FLAG_BE_LIBERAL
;
1146 /* tcp_packet will set them */
1147 ct
->proto
.tcp
.last_index
= TCP_NONE_SET
;
1149 pr_debug("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i "
1150 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
1151 sender
->td_end
, sender
->td_maxend
, sender
->td_maxwin
,
1153 receiver
->td_end
, receiver
->td_maxend
, receiver
->td_maxwin
,
1154 receiver
->td_scale
);
1158 static bool tcp_can_early_drop(const struct nf_conn
*ct
)
1160 switch (ct
->proto
.tcp
.state
) {
1161 case TCP_CONNTRACK_FIN_WAIT
:
1162 case TCP_CONNTRACK_LAST_ACK
:
1163 case TCP_CONNTRACK_TIME_WAIT
:
1164 case TCP_CONNTRACK_CLOSE
:
1165 case TCP_CONNTRACK_CLOSE_WAIT
:
1174 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1176 #include <linux/netfilter/nfnetlink.h>
1177 #include <linux/netfilter/nfnetlink_conntrack.h>
1179 static int tcp_to_nlattr(struct sk_buff
*skb
, struct nlattr
*nla
,
1182 struct nlattr
*nest_parms
;
1183 struct nf_ct_tcp_flags tmp
= {};
1185 spin_lock_bh(&ct
->lock
);
1186 nest_parms
= nla_nest_start(skb
, CTA_PROTOINFO_TCP
| NLA_F_NESTED
);
1188 goto nla_put_failure
;
1190 if (nla_put_u8(skb
, CTA_PROTOINFO_TCP_STATE
, ct
->proto
.tcp
.state
) ||
1191 nla_put_u8(skb
, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL
,
1192 ct
->proto
.tcp
.seen
[0].td_scale
) ||
1193 nla_put_u8(skb
, CTA_PROTOINFO_TCP_WSCALE_REPLY
,
1194 ct
->proto
.tcp
.seen
[1].td_scale
))
1195 goto nla_put_failure
;
1197 tmp
.flags
= ct
->proto
.tcp
.seen
[0].flags
;
1198 if (nla_put(skb
, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL
,
1199 sizeof(struct nf_ct_tcp_flags
), &tmp
))
1200 goto nla_put_failure
;
1202 tmp
.flags
= ct
->proto
.tcp
.seen
[1].flags
;
1203 if (nla_put(skb
, CTA_PROTOINFO_TCP_FLAGS_REPLY
,
1204 sizeof(struct nf_ct_tcp_flags
), &tmp
))
1205 goto nla_put_failure
;
1206 spin_unlock_bh(&ct
->lock
);
1208 nla_nest_end(skb
, nest_parms
);
1213 spin_unlock_bh(&ct
->lock
);
1217 static const struct nla_policy tcp_nla_policy
[CTA_PROTOINFO_TCP_MAX
+1] = {
1218 [CTA_PROTOINFO_TCP_STATE
] = { .type
= NLA_U8
},
1219 [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL
] = { .type
= NLA_U8
},
1220 [CTA_PROTOINFO_TCP_WSCALE_REPLY
] = { .type
= NLA_U8
},
1221 [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL
] = { .len
= sizeof(struct nf_ct_tcp_flags
) },
1222 [CTA_PROTOINFO_TCP_FLAGS_REPLY
] = { .len
= sizeof(struct nf_ct_tcp_flags
) },
1225 static int nlattr_to_tcp(struct nlattr
*cda
[], struct nf_conn
*ct
)
1227 struct nlattr
*pattr
= cda
[CTA_PROTOINFO_TCP
];
1228 struct nlattr
*tb
[CTA_PROTOINFO_TCP_MAX
+1];
1231 /* updates could not contain anything about the private
1232 * protocol info, in that case skip the parsing */
1236 err
= nla_parse_nested(tb
, CTA_PROTOINFO_TCP_MAX
, pattr
,
1237 tcp_nla_policy
, NULL
);
1241 if (tb
[CTA_PROTOINFO_TCP_STATE
] &&
1242 nla_get_u8(tb
[CTA_PROTOINFO_TCP_STATE
]) >= TCP_CONNTRACK_MAX
)
1245 spin_lock_bh(&ct
->lock
);
1246 if (tb
[CTA_PROTOINFO_TCP_STATE
])
1247 ct
->proto
.tcp
.state
= nla_get_u8(tb
[CTA_PROTOINFO_TCP_STATE
]);
1249 if (tb
[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL
]) {
1250 struct nf_ct_tcp_flags
*attr
=
1251 nla_data(tb
[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL
]);
1252 ct
->proto
.tcp
.seen
[0].flags
&= ~attr
->mask
;
1253 ct
->proto
.tcp
.seen
[0].flags
|= attr
->flags
& attr
->mask
;
1256 if (tb
[CTA_PROTOINFO_TCP_FLAGS_REPLY
]) {
1257 struct nf_ct_tcp_flags
*attr
=
1258 nla_data(tb
[CTA_PROTOINFO_TCP_FLAGS_REPLY
]);
1259 ct
->proto
.tcp
.seen
[1].flags
&= ~attr
->mask
;
1260 ct
->proto
.tcp
.seen
[1].flags
|= attr
->flags
& attr
->mask
;
1263 if (tb
[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL
] &&
1264 tb
[CTA_PROTOINFO_TCP_WSCALE_REPLY
] &&
1265 ct
->proto
.tcp
.seen
[0].flags
& IP_CT_TCP_FLAG_WINDOW_SCALE
&&
1266 ct
->proto
.tcp
.seen
[1].flags
& IP_CT_TCP_FLAG_WINDOW_SCALE
) {
1267 ct
->proto
.tcp
.seen
[0].td_scale
=
1268 nla_get_u8(tb
[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL
]);
1269 ct
->proto
.tcp
.seen
[1].td_scale
=
1270 nla_get_u8(tb
[CTA_PROTOINFO_TCP_WSCALE_REPLY
]);
1272 spin_unlock_bh(&ct
->lock
);
1277 static int tcp_nlattr_size(void)
1279 return nla_total_size(0) /* CTA_PROTOINFO_TCP */
1280 + nla_policy_len(tcp_nla_policy
, CTA_PROTOINFO_TCP_MAX
+ 1);
1283 static unsigned int tcp_nlattr_tuple_size(void)
1285 static unsigned int size __read_mostly
;
1288 size
= nla_policy_len(nf_ct_port_nla_policy
, CTA_PROTO_MAX
+ 1);
1294 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1296 #include <linux/netfilter/nfnetlink.h>
1297 #include <linux/netfilter/nfnetlink_cttimeout.h>
1299 static int tcp_timeout_nlattr_to_obj(struct nlattr
*tb
[],
1300 struct net
*net
, void *data
)
1302 unsigned int *timeouts
= data
;
1303 struct nf_tcp_net
*tn
= tcp_pernet(net
);
1306 /* set default TCP timeouts. */
1307 for (i
=0; i
<TCP_CONNTRACK_TIMEOUT_MAX
; i
++)
1308 timeouts
[i
] = tn
->timeouts
[i
];
1310 if (tb
[CTA_TIMEOUT_TCP_SYN_SENT
]) {
1311 timeouts
[TCP_CONNTRACK_SYN_SENT
] =
1312 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_SYN_SENT
]))*HZ
;
1314 if (tb
[CTA_TIMEOUT_TCP_SYN_RECV
]) {
1315 timeouts
[TCP_CONNTRACK_SYN_RECV
] =
1316 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_SYN_RECV
]))*HZ
;
1318 if (tb
[CTA_TIMEOUT_TCP_ESTABLISHED
]) {
1319 timeouts
[TCP_CONNTRACK_ESTABLISHED
] =
1320 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_ESTABLISHED
]))*HZ
;
1322 if (tb
[CTA_TIMEOUT_TCP_FIN_WAIT
]) {
1323 timeouts
[TCP_CONNTRACK_FIN_WAIT
] =
1324 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_FIN_WAIT
]))*HZ
;
1326 if (tb
[CTA_TIMEOUT_TCP_CLOSE_WAIT
]) {
1327 timeouts
[TCP_CONNTRACK_CLOSE_WAIT
] =
1328 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_CLOSE_WAIT
]))*HZ
;
1330 if (tb
[CTA_TIMEOUT_TCP_LAST_ACK
]) {
1331 timeouts
[TCP_CONNTRACK_LAST_ACK
] =
1332 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_LAST_ACK
]))*HZ
;
1334 if (tb
[CTA_TIMEOUT_TCP_TIME_WAIT
]) {
1335 timeouts
[TCP_CONNTRACK_TIME_WAIT
] =
1336 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_TIME_WAIT
]))*HZ
;
1338 if (tb
[CTA_TIMEOUT_TCP_CLOSE
]) {
1339 timeouts
[TCP_CONNTRACK_CLOSE
] =
1340 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_CLOSE
]))*HZ
;
1342 if (tb
[CTA_TIMEOUT_TCP_SYN_SENT2
]) {
1343 timeouts
[TCP_CONNTRACK_SYN_SENT2
] =
1344 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_SYN_SENT2
]))*HZ
;
1346 if (tb
[CTA_TIMEOUT_TCP_RETRANS
]) {
1347 timeouts
[TCP_CONNTRACK_RETRANS
] =
1348 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_RETRANS
]))*HZ
;
1350 if (tb
[CTA_TIMEOUT_TCP_UNACK
]) {
1351 timeouts
[TCP_CONNTRACK_UNACK
] =
1352 ntohl(nla_get_be32(tb
[CTA_TIMEOUT_TCP_UNACK
]))*HZ
;
1358 tcp_timeout_obj_to_nlattr(struct sk_buff
*skb
, const void *data
)
1360 const unsigned int *timeouts
= data
;
1362 if (nla_put_be32(skb
, CTA_TIMEOUT_TCP_SYN_SENT
,
1363 htonl(timeouts
[TCP_CONNTRACK_SYN_SENT
] / HZ
)) ||
1364 nla_put_be32(skb
, CTA_TIMEOUT_TCP_SYN_RECV
,
1365 htonl(timeouts
[TCP_CONNTRACK_SYN_RECV
] / HZ
)) ||
1366 nla_put_be32(skb
, CTA_TIMEOUT_TCP_ESTABLISHED
,
1367 htonl(timeouts
[TCP_CONNTRACK_ESTABLISHED
] / HZ
)) ||
1368 nla_put_be32(skb
, CTA_TIMEOUT_TCP_FIN_WAIT
,
1369 htonl(timeouts
[TCP_CONNTRACK_FIN_WAIT
] / HZ
)) ||
1370 nla_put_be32(skb
, CTA_TIMEOUT_TCP_CLOSE_WAIT
,
1371 htonl(timeouts
[TCP_CONNTRACK_CLOSE_WAIT
] / HZ
)) ||
1372 nla_put_be32(skb
, CTA_TIMEOUT_TCP_LAST_ACK
,
1373 htonl(timeouts
[TCP_CONNTRACK_LAST_ACK
] / HZ
)) ||
1374 nla_put_be32(skb
, CTA_TIMEOUT_TCP_TIME_WAIT
,
1375 htonl(timeouts
[TCP_CONNTRACK_TIME_WAIT
] / HZ
)) ||
1376 nla_put_be32(skb
, CTA_TIMEOUT_TCP_CLOSE
,
1377 htonl(timeouts
[TCP_CONNTRACK_CLOSE
] / HZ
)) ||
1378 nla_put_be32(skb
, CTA_TIMEOUT_TCP_SYN_SENT2
,
1379 htonl(timeouts
[TCP_CONNTRACK_SYN_SENT2
] / HZ
)) ||
1380 nla_put_be32(skb
, CTA_TIMEOUT_TCP_RETRANS
,
1381 htonl(timeouts
[TCP_CONNTRACK_RETRANS
] / HZ
)) ||
1382 nla_put_be32(skb
, CTA_TIMEOUT_TCP_UNACK
,
1383 htonl(timeouts
[TCP_CONNTRACK_UNACK
] / HZ
)))
1384 goto nla_put_failure
;
1391 static const struct nla_policy tcp_timeout_nla_policy
[CTA_TIMEOUT_TCP_MAX
+1] = {
1392 [CTA_TIMEOUT_TCP_SYN_SENT
] = { .type
= NLA_U32
},
1393 [CTA_TIMEOUT_TCP_SYN_RECV
] = { .type
= NLA_U32
},
1394 [CTA_TIMEOUT_TCP_ESTABLISHED
] = { .type
= NLA_U32
},
1395 [CTA_TIMEOUT_TCP_FIN_WAIT
] = { .type
= NLA_U32
},
1396 [CTA_TIMEOUT_TCP_CLOSE_WAIT
] = { .type
= NLA_U32
},
1397 [CTA_TIMEOUT_TCP_LAST_ACK
] = { .type
= NLA_U32
},
1398 [CTA_TIMEOUT_TCP_TIME_WAIT
] = { .type
= NLA_U32
},
1399 [CTA_TIMEOUT_TCP_CLOSE
] = { .type
= NLA_U32
},
1400 [CTA_TIMEOUT_TCP_SYN_SENT2
] = { .type
= NLA_U32
},
1401 [CTA_TIMEOUT_TCP_RETRANS
] = { .type
= NLA_U32
},
1402 [CTA_TIMEOUT_TCP_UNACK
] = { .type
= NLA_U32
},
1404 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1406 #ifdef CONFIG_SYSCTL
1407 static struct ctl_table tcp_sysctl_table
[] = {
1409 .procname
= "nf_conntrack_tcp_timeout_syn_sent",
1410 .maxlen
= sizeof(unsigned int),
1412 .proc_handler
= proc_dointvec_jiffies
,
1415 .procname
= "nf_conntrack_tcp_timeout_syn_recv",
1416 .maxlen
= sizeof(unsigned int),
1418 .proc_handler
= proc_dointvec_jiffies
,
1421 .procname
= "nf_conntrack_tcp_timeout_established",
1422 .maxlen
= sizeof(unsigned int),
1424 .proc_handler
= proc_dointvec_jiffies
,
1427 .procname
= "nf_conntrack_tcp_timeout_fin_wait",
1428 .maxlen
= sizeof(unsigned int),
1430 .proc_handler
= proc_dointvec_jiffies
,
1433 .procname
= "nf_conntrack_tcp_timeout_close_wait",
1434 .maxlen
= sizeof(unsigned int),
1436 .proc_handler
= proc_dointvec_jiffies
,
1439 .procname
= "nf_conntrack_tcp_timeout_last_ack",
1440 .maxlen
= sizeof(unsigned int),
1442 .proc_handler
= proc_dointvec_jiffies
,
1445 .procname
= "nf_conntrack_tcp_timeout_time_wait",
1446 .maxlen
= sizeof(unsigned int),
1448 .proc_handler
= proc_dointvec_jiffies
,
1451 .procname
= "nf_conntrack_tcp_timeout_close",
1452 .maxlen
= sizeof(unsigned int),
1454 .proc_handler
= proc_dointvec_jiffies
,
1457 .procname
= "nf_conntrack_tcp_timeout_max_retrans",
1458 .maxlen
= sizeof(unsigned int),
1460 .proc_handler
= proc_dointvec_jiffies
,
1463 .procname
= "nf_conntrack_tcp_timeout_unacknowledged",
1464 .maxlen
= sizeof(unsigned int),
1466 .proc_handler
= proc_dointvec_jiffies
,
1469 .procname
= "nf_conntrack_tcp_loose",
1470 .maxlen
= sizeof(unsigned int),
1472 .proc_handler
= proc_dointvec
,
1475 .procname
= "nf_conntrack_tcp_be_liberal",
1476 .maxlen
= sizeof(unsigned int),
1478 .proc_handler
= proc_dointvec
,
1481 .procname
= "nf_conntrack_tcp_max_retrans",
1482 .maxlen
= sizeof(unsigned int),
1484 .proc_handler
= proc_dointvec
,
1488 #endif /* CONFIG_SYSCTL */
1490 static int tcp_kmemdup_sysctl_table(struct nf_proto_net
*pn
,
1491 struct nf_tcp_net
*tn
)
1493 #ifdef CONFIG_SYSCTL
1497 pn
->ctl_table
= kmemdup(tcp_sysctl_table
,
1498 sizeof(tcp_sysctl_table
),
1503 pn
->ctl_table
[0].data
= &tn
->timeouts
[TCP_CONNTRACK_SYN_SENT
];
1504 pn
->ctl_table
[1].data
= &tn
->timeouts
[TCP_CONNTRACK_SYN_RECV
];
1505 pn
->ctl_table
[2].data
= &tn
->timeouts
[TCP_CONNTRACK_ESTABLISHED
];
1506 pn
->ctl_table
[3].data
= &tn
->timeouts
[TCP_CONNTRACK_FIN_WAIT
];
1507 pn
->ctl_table
[4].data
= &tn
->timeouts
[TCP_CONNTRACK_CLOSE_WAIT
];
1508 pn
->ctl_table
[5].data
= &tn
->timeouts
[TCP_CONNTRACK_LAST_ACK
];
1509 pn
->ctl_table
[6].data
= &tn
->timeouts
[TCP_CONNTRACK_TIME_WAIT
];
1510 pn
->ctl_table
[7].data
= &tn
->timeouts
[TCP_CONNTRACK_CLOSE
];
1511 pn
->ctl_table
[8].data
= &tn
->timeouts
[TCP_CONNTRACK_RETRANS
];
1512 pn
->ctl_table
[9].data
= &tn
->timeouts
[TCP_CONNTRACK_UNACK
];
1513 pn
->ctl_table
[10].data
= &tn
->tcp_loose
;
1514 pn
->ctl_table
[11].data
= &tn
->tcp_be_liberal
;
1515 pn
->ctl_table
[12].data
= &tn
->tcp_max_retrans
;
1520 static int tcp_init_net(struct net
*net
, u_int16_t proto
)
1522 struct nf_tcp_net
*tn
= tcp_pernet(net
);
1523 struct nf_proto_net
*pn
= &tn
->pn
;
1528 for (i
= 0; i
< TCP_CONNTRACK_TIMEOUT_MAX
; i
++)
1529 tn
->timeouts
[i
] = tcp_timeouts
[i
];
1531 tn
->tcp_loose
= nf_ct_tcp_loose
;
1532 tn
->tcp_be_liberal
= nf_ct_tcp_be_liberal
;
1533 tn
->tcp_max_retrans
= nf_ct_tcp_max_retrans
;
1536 return tcp_kmemdup_sysctl_table(pn
, tn
);
1539 static struct nf_proto_net
*tcp_get_net_proto(struct net
*net
)
1541 return &net
->ct
.nf_ct_proto
.tcp
.pn
;
1544 struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp4 __read_mostly
=
1547 .l4proto
= IPPROTO_TCP
,
1548 .pkt_to_tuple
= tcp_pkt_to_tuple
,
1549 .invert_tuple
= tcp_invert_tuple
,
1550 #ifdef CONFIG_NF_CONNTRACK_PROCFS
1551 .print_conntrack
= tcp_print_conntrack
,
1553 .packet
= tcp_packet
,
1554 .get_timeouts
= tcp_get_timeouts
,
1557 .can_early_drop
= tcp_can_early_drop
,
1558 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1559 .to_nlattr
= tcp_to_nlattr
,
1560 .nlattr_size
= tcp_nlattr_size
,
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
,
1564 .nlattr_tuple_size
= tcp_nlattr_tuple_size
,
1565 .nla_policy
= nf_ct_port_nla_policy
,
1567 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_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
,
1576 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1577 .init_net
= tcp_init_net
,
1578 .get_net_proto
= tcp_get_net_proto
,
1580 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp4
);
1582 struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6 __read_mostly
=
1584 .l3proto
= PF_INET6
,
1585 .l4proto
= IPPROTO_TCP
,
1586 .pkt_to_tuple
= tcp_pkt_to_tuple
,
1587 .invert_tuple
= tcp_invert_tuple
,
1588 #ifdef CONFIG_NF_CONNTRACK_PROCFS
1589 .print_conntrack
= tcp_print_conntrack
,
1591 .packet
= tcp_packet
,
1592 .get_timeouts
= tcp_get_timeouts
,
1595 .can_early_drop
= tcp_can_early_drop
,
1596 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1597 .to_nlattr
= tcp_to_nlattr
,
1598 .nlattr_size
= tcp_nlattr_size
,
1599 .from_nlattr
= nlattr_to_tcp
,
1600 .tuple_to_nlattr
= nf_ct_port_tuple_to_nlattr
,
1601 .nlattr_to_tuple
= nf_ct_port_nlattr_to_tuple
,
1602 .nlattr_tuple_size
= tcp_nlattr_tuple_size
,
1603 .nla_policy
= nf_ct_port_nla_policy
,
1605 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1607 .nlattr_to_obj
= tcp_timeout_nlattr_to_obj
,
1608 .obj_to_nlattr
= tcp_timeout_obj_to_nlattr
,
1609 .nlattr_max
= CTA_TIMEOUT_TCP_MAX
,
1610 .obj_size
= sizeof(unsigned int) *
1611 TCP_CONNTRACK_TIMEOUT_MAX
,
1612 .nla_policy
= tcp_timeout_nla_policy
,
1614 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1615 .init_net
= tcp_init_net
,
1616 .get_net_proto
= tcp_get_net_proto
,
1618 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp6
);