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 8 bytes. */
286 hp = skb_header_pointer(skb, dataoff, 8, &_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 /* Print out the per-protocol part of the tuple. */
305 static void tcp_print_tuple(struct seq_file *s,
306 const struct nf_conntrack_tuple *tuple)
308 seq_printf(s, "sport=%hu dport=%hu ",
309 ntohs(tuple->src.u.tcp.port),
310 ntohs(tuple->dst.u.tcp.port));
313 /* Print out the private part of the conntrack. */
314 static void tcp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
316 enum tcp_conntrack state;
318 spin_lock_bh(&ct->lock);
319 state = ct->proto.tcp.state;
320 spin_unlock_bh(&ct->lock);
322 seq_printf(s, "%s ", tcp_conntrack_names[state]);
325 static unsigned int get_conntrack_index(const struct tcphdr *tcph)
327 if (tcph->rst) return TCP_RST_SET;
328 else if (tcph->syn) return (tcph->ack ? TCP_SYNACK_SET : TCP_SYN_SET);
329 else if (tcph->fin) return TCP_FIN_SET;
330 else if (tcph->ack) return TCP_ACK_SET;
331 else return TCP_NONE_SET;
334 /* TCP connection tracking based on 'Real Stateful TCP Packet Filtering
335 in IP Filter' by Guido van Rooij.
337 http://www.sane.nl/events/sane2000/papers.html
338 http://www.darkart.com/mirrors/www.obfuscation.org/ipf/
340 The boundaries and the conditions are changed according to RFC793:
341 the packet must intersect the window (i.e. segments may be
342 after the right or before the left edge) and thus receivers may ACK
343 segments after the right edge of the window.
345 td_maxend = max(sack + max(win,1)) seen in reply packets
346 td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets
347 td_maxwin += seq + len - sender.td_maxend
348 if seq + len > sender.td_maxend
349 td_end = max(seq + len) seen in sent packets
351 I. Upper bound for valid data: seq <= sender.td_maxend
352 II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin
353 III. Upper bound for valid (s)ack: sack <= receiver.td_end
354 IV. Lower bound for valid (s)ack: sack >= receiver.td_end - MAXACKWINDOW
356 where sack is the highest right edge of sack block found in the packet
357 or ack in the case of packet without SACK option.
359 The upper bound limit for a valid (s)ack is not ignored -
360 we doesn't have to deal with fragments.
363 static inline __u32 segment_seq_plus_len(__u32 seq,
365 unsigned int dataoff,
366 const struct tcphdr *tcph)
368 /* XXX Should I use payload length field in IP/IPv6 header ?
370 return (seq + len - dataoff - tcph->doff*4
371 + (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0));
374 /* Fixme: what about big packets? */
375 #define MAXACKWINCONST 66000
376 #define MAXACKWINDOW(sender) \
377 ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \
381 * Simplified tcp_parse_options routine from tcp_input.c
383 static void tcp_options(const struct sk_buff *skb,
384 unsigned int dataoff,
385 const struct tcphdr *tcph,
386 struct ip_ct_tcp_state *state)
388 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
389 const unsigned char *ptr;
390 int length = (tcph->doff*4) - sizeof(struct tcphdr);
395 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
409 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
416 if (opsize < 2) /* "silly options" */
419 return; /* don't parse partial options */
421 if (opcode == TCPOPT_SACK_PERM
422 && opsize == TCPOLEN_SACK_PERM)
423 state->flags |= IP_CT_TCP_FLAG_SACK_PERM;
424 else if (opcode == TCPOPT_WINDOW
425 && opsize == TCPOLEN_WINDOW) {
426 state->td_scale = *(u_int8_t *)ptr;
428 if (state->td_scale > 14) {
430 state->td_scale = 14;
433 IP_CT_TCP_FLAG_WINDOW_SCALE;
441 static void tcp_sack(const struct sk_buff *skb, unsigned int dataoff,
442 const struct tcphdr *tcph, __u32 *sack)
444 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
445 const unsigned char *ptr;
446 int length = (tcph->doff*4) - sizeof(struct tcphdr);
452 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
456 /* Fast path for timestamp-only option */
457 if (length == TCPOLEN_TSTAMP_ALIGNED
458 && *(__be32 *)ptr == htonl((TCPOPT_NOP << 24)
460 | (TCPOPT_TIMESTAMP << 8)
461 | TCPOLEN_TIMESTAMP))
471 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
478 if (opsize < 2) /* "silly options" */
481 return; /* don't parse partial options */
483 if (opcode == TCPOPT_SACK
484 && opsize >= (TCPOLEN_SACK_BASE
485 + TCPOLEN_SACK_PERBLOCK)
486 && !((opsize - TCPOLEN_SACK_BASE)
487 % TCPOLEN_SACK_PERBLOCK)) {
489 i < (opsize - TCPOLEN_SACK_BASE);
490 i += TCPOLEN_SACK_PERBLOCK) {
491 tmp = get_unaligned_be32((__be32 *)(ptr+i)+1);
493 if (after(tmp, *sack))
504 static bool tcp_in_window(const struct nf_conn *ct,
505 struct ip_ct_tcp *state,
506 enum ip_conntrack_dir dir,
508 const struct sk_buff *skb,
509 unsigned int dataoff,
510 const struct tcphdr *tcph,
513 struct net *net = nf_ct_net(ct);
514 struct nf_tcp_net *tn = tcp_pernet(net);
515 struct ip_ct_tcp_state *sender = &state->seen[dir];
516 struct ip_ct_tcp_state *receiver = &state->seen[!dir];
517 const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
518 __u32 seq, ack, sack, end, win, swin;
520 bool res, in_recv_win;
523 * Get the required data from the packet.
525 seq = ntohl(tcph->seq);
526 ack = sack = ntohl(tcph->ack_seq);
527 win = ntohs(tcph->window);
528 end = segment_seq_plus_len(seq, skb->len, dataoff, tcph);
530 if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM)
531 tcp_sack(skb, dataoff, tcph, &sack);
533 /* Take into account NAT sequence number mangling */
534 receiver_offset = nf_ct_seq_offset(ct, !dir, ack - 1);
535 ack -= receiver_offset;
536 sack -= receiver_offset;
538 pr_debug("tcp_in_window: START\n");
539 pr_debug("tcp_in_window: ");
540 nf_ct_dump_tuple(tuple);
541 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
542 seq, ack, receiver_offset, sack, receiver_offset, win, end);
543 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
544 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
545 sender->td_end, sender->td_maxend, sender->td_maxwin,
547 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
550 if (sender->td_maxwin == 0) {
552 * Initialize sender data.
556 * SYN-ACK in reply to a SYN
557 * or SYN from reply direction in simultaneous open.
560 sender->td_maxend = end;
561 sender->td_maxwin = (win == 0 ? 1 : win);
563 tcp_options(skb, dataoff, tcph, sender);
566 * Both sides must send the Window Scale option
567 * to enable window scaling in either direction.
569 if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE
570 && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE))
572 receiver->td_scale = 0;
574 /* Simultaneous open */
578 * We are in the middle of a connection,
579 * its history is lost for us.
580 * Let's try to use the data from the packet.
582 sender->td_end = end;
583 swin = win << sender->td_scale;
584 sender->td_maxwin = (swin == 0 ? 1 : swin);
585 sender->td_maxend = end + sender->td_maxwin;
587 * We haven't seen traffic in the other direction yet
588 * but we have to tweak window tracking to pass III
589 * and IV until that happens.
591 if (receiver->td_maxwin == 0)
592 receiver->td_end = receiver->td_maxend = sack;
594 } else if (((state->state == TCP_CONNTRACK_SYN_SENT
595 && dir == IP_CT_DIR_ORIGINAL)
596 || (state->state == TCP_CONNTRACK_SYN_RECV
597 && dir == IP_CT_DIR_REPLY))
598 && after(end, sender->td_end)) {
600 * RFC 793: "if a TCP is reinitialized ... then it need
601 * not wait at all; it must only be sure to use sequence
602 * numbers larger than those recently used."
605 sender->td_maxend = end;
606 sender->td_maxwin = (win == 0 ? 1 : win);
608 tcp_options(skb, dataoff, tcph, sender);
613 * If there is no ACK, just pretend it was set and OK.
615 ack = sack = receiver->td_end;
616 } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) ==
617 (TCP_FLAG_ACK|TCP_FLAG_RST))
620 * Broken TCP stacks, that set ACK in RST packets as well
621 * with zero ack value.
623 ack = sack = receiver->td_end;
626 if (tcph->rst && seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT)
628 * RST sent answering SYN.
630 seq = end = sender->td_end;
632 pr_debug("tcp_in_window: ");
633 nf_ct_dump_tuple(tuple);
634 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
635 seq, ack, receiver_offset, sack, receiver_offset, win, end);
636 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
637 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
638 sender->td_end, sender->td_maxend, sender->td_maxwin,
640 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
643 /* Is the ending sequence in the receive window (if available)? */
644 in_recv_win = !receiver->td_maxwin ||
645 after(end, sender->td_end - receiver->td_maxwin - 1);
647 pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
648 before(seq, sender->td_maxend + 1),
649 (in_recv_win ? 1 : 0),
650 before(sack, receiver->td_end + 1),
651 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1));
653 if (before(seq, sender->td_maxend + 1) &&
655 before(sack, receiver->td_end + 1) &&
656 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)) {
658 * Take into account window scaling (RFC 1323).
661 win <<= sender->td_scale;
664 * Update sender data.
666 swin = win + (sack - ack);
667 if (sender->td_maxwin < swin)
668 sender->td_maxwin = swin;
669 if (after(end, sender->td_end)) {
670 sender->td_end = end;
671 sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
674 if (!(sender->flags & IP_CT_TCP_FLAG_MAXACK_SET)) {
675 sender->td_maxack = ack;
676 sender->flags |= IP_CT_TCP_FLAG_MAXACK_SET;
677 } else if (after(ack, sender->td_maxack))
678 sender->td_maxack = ack;
682 * Update receiver data.
684 if (receiver->td_maxwin != 0 && after(end, sender->td_maxend))
685 receiver->td_maxwin += end - sender->td_maxend;
686 if (after(sack + win, receiver->td_maxend - 1)) {
687 receiver->td_maxend = sack + win;
689 receiver->td_maxend++;
691 if (ack == receiver->td_end)
692 receiver->flags &= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
695 * Check retransmissions.
697 if (index == TCP_ACK_SET) {
698 if (state->last_dir == dir
699 && state->last_seq == seq
700 && state->last_ack == ack
701 && state->last_end == end
702 && state->last_win == win)
705 state->last_dir = dir;
706 state->last_seq = seq;
707 state->last_ack = ack;
708 state->last_end = end;
709 state->last_win = win;
716 if (sender->flags & IP_CT_TCP_FLAG_BE_LIBERAL ||
719 if (!res && LOG_INVALID(net, IPPROTO_TCP))
720 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
722 before(seq, sender->td_maxend + 1) ?
724 before(sack, receiver->td_end + 1) ?
725 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1) ? "BUG"
726 : "ACK is under the lower bound (possible overly delayed ACK)"
727 : "ACK is over the upper bound (ACKed data not seen yet)"
728 : "SEQ is under the lower bound (already ACKed data retransmitted)"
729 : "SEQ is over the upper bound (over the window of the receiver)");
732 pr_debug("tcp_in_window: res=%u sender end=%u maxend=%u maxwin=%u "
733 "receiver end=%u maxend=%u maxwin=%u\n",
734 res, sender->td_end, sender->td_maxend, sender->td_maxwin,
735 receiver->td_end, receiver->td_maxend, receiver->td_maxwin);
740 /* table of valid flag combinations - PUSH, ECE and CWR are always valid */
741 static const u8 tcp_valid_flags[(TCPHDR_FIN|TCPHDR_SYN|TCPHDR_RST|TCPHDR_ACK|
745 [TCPHDR_SYN|TCPHDR_URG] = 1,
746 [TCPHDR_SYN|TCPHDR_ACK] = 1,
748 [TCPHDR_RST|TCPHDR_ACK] = 1,
749 [TCPHDR_FIN|TCPHDR_ACK] = 1,
750 [TCPHDR_FIN|TCPHDR_ACK|TCPHDR_URG] = 1,
752 [TCPHDR_ACK|TCPHDR_URG] = 1,
755 /* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
756 static int tcp_error(struct net *net, struct nf_conn *tmpl,
758 unsigned int dataoff,
759 enum ip_conntrack_info *ctinfo,
761 unsigned int hooknum)
763 const struct tcphdr *th;
765 unsigned int tcplen = skb->len - dataoff;
768 /* Smaller that minimal TCP header? */
769 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
771 if (LOG_INVALID(net, IPPROTO_TCP))
772 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
773 "nf_ct_tcp: short packet ");
777 /* Not whole TCP header or malformed packet */
778 if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) {
779 if (LOG_INVALID(net, IPPROTO_TCP))
780 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
781 "nf_ct_tcp: truncated/malformed packet ");
785 /* Checksum invalid? Ignore.
786 * We skip checking packets on the outgoing path
787 * because the checksum is assumed to be correct.
789 /* FIXME: Source route IP option packets --RR */
790 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
791 nf_checksum(skb, hooknum, dataoff, IPPROTO_TCP, pf)) {
792 if (LOG_INVALID(net, IPPROTO_TCP))
793 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
794 "nf_ct_tcp: bad TCP checksum ");
798 /* Check TCP flags. */
799 tcpflags = (tcp_flag_byte(th) & ~(TCPHDR_ECE|TCPHDR_CWR|TCPHDR_PSH));
800 if (!tcp_valid_flags[tcpflags]) {
801 if (LOG_INVALID(net, IPPROTO_TCP))
802 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
803 "nf_ct_tcp: invalid TCP flag combination ");
810 static unsigned int *tcp_get_timeouts(struct net *net)
812 return tcp_pernet(net)->timeouts;
815 /* Returns verdict for packet, or -1 for invalid. */
816 static int tcp_packet(struct nf_conn *ct,
817 const struct sk_buff *skb,
818 unsigned int dataoff,
819 enum ip_conntrack_info ctinfo,
821 unsigned int hooknum,
822 unsigned int *timeouts)
824 struct net *net = nf_ct_net(ct);
825 struct nf_tcp_net *tn = tcp_pernet(net);
826 struct nf_conntrack_tuple *tuple;
827 enum tcp_conntrack new_state, old_state;
828 enum ip_conntrack_dir dir;
829 const struct tcphdr *th;
831 unsigned long timeout;
834 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
837 spin_lock_bh(&ct->lock);
838 old_state = ct->proto.tcp.state;
839 dir = CTINFO2DIR(ctinfo);
840 index = get_conntrack_index(th);
841 new_state = tcp_conntracks[dir][index][old_state];
842 tuple = &ct->tuplehash[dir].tuple;
845 case TCP_CONNTRACK_SYN_SENT:
846 if (old_state < TCP_CONNTRACK_TIME_WAIT)
848 /* RFC 1122: "When a connection is closed actively,
849 * it MUST linger in TIME-WAIT state for a time 2xMSL
850 * (Maximum Segment Lifetime). However, it MAY accept
851 * a new SYN from the remote TCP to reopen the connection
852 * directly from TIME-WAIT state, if..."
853 * We ignore the conditions because we are in the
854 * TIME-WAIT state anyway.
856 * Handle aborted connections: we and the server
857 * think there is an existing connection but the client
858 * aborts it and starts a new one.
860 if (((ct->proto.tcp.seen[dir].flags
861 | ct->proto.tcp.seen[!dir].flags)
862 & IP_CT_TCP_FLAG_CLOSE_INIT)
863 || (ct->proto.tcp.last_dir == dir
864 && ct->proto.tcp.last_index == TCP_RST_SET)) {
865 /* Attempt to reopen a closed/aborted connection.
866 * Delete this connection and look up again. */
867 spin_unlock_bh(&ct->lock);
869 /* Only repeat if we can actually remove the timer.
870 * Destruction may already be in progress in process
871 * context and we must give it a chance to terminate.
878 case TCP_CONNTRACK_IGNORE:
881 * Our connection entry may be out of sync, so ignore
882 * packets which may signal the real connection between
883 * the client and the server.
886 * b) SYN/ACK in REPLY
887 * c) ACK in reply direction after initial SYN in original.
889 * If the ignored packet is invalid, the receiver will send
890 * a RST we'll catch below.
892 if (index == TCP_SYNACK_SET
893 && ct->proto.tcp.last_index == TCP_SYN_SET
894 && ct->proto.tcp.last_dir != dir
895 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
896 /* b) This SYN/ACK acknowledges a SYN that we earlier
897 * ignored as invalid. This means that the client and
898 * the server are both in sync, while the firewall is
899 * not. We get in sync from the previously annotated
902 old_state = TCP_CONNTRACK_SYN_SENT;
903 new_state = TCP_CONNTRACK_SYN_RECV;
904 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_end =
905 ct->proto.tcp.last_end;
906 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxend =
907 ct->proto.tcp.last_end;
908 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxwin =
909 ct->proto.tcp.last_win == 0 ?
910 1 : ct->proto.tcp.last_win;
911 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_scale =
912 ct->proto.tcp.last_wscale;
913 ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK;
914 ct->proto.tcp.seen[ct->proto.tcp.last_dir].flags =
915 ct->proto.tcp.last_flags;
916 memset(&ct->proto.tcp.seen[dir], 0,
917 sizeof(struct ip_ct_tcp_state));
920 ct->proto.tcp.last_index = index;
921 ct->proto.tcp.last_dir = dir;
922 ct->proto.tcp.last_seq = ntohl(th->seq);
923 ct->proto.tcp.last_end =
924 segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th);
925 ct->proto.tcp.last_win = ntohs(th->window);
927 /* a) This is a SYN in ORIGINAL. The client and the server
928 * may be in sync but we are not. In that case, we annotate
929 * the TCP options and let the packet go through. If it is a
930 * valid SYN packet, the server will reply with a SYN/ACK, and
931 * then we'll get in sync. Otherwise, the server potentially
932 * responds with a challenge ACK if implementing RFC5961.
934 if (index == TCP_SYN_SET && dir == IP_CT_DIR_ORIGINAL) {
935 struct ip_ct_tcp_state seen = {};
937 ct->proto.tcp.last_flags =
938 ct->proto.tcp.last_wscale = 0;
939 tcp_options(skb, dataoff, th, &seen);
940 if (seen.flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
941 ct->proto.tcp.last_flags |=
942 IP_CT_TCP_FLAG_WINDOW_SCALE;
943 ct->proto.tcp.last_wscale = seen.td_scale;
945 if (seen.flags & IP_CT_TCP_FLAG_SACK_PERM) {
946 ct->proto.tcp.last_flags |=
947 IP_CT_TCP_FLAG_SACK_PERM;
949 /* Mark the potential for RFC5961 challenge ACK,
950 * this pose a special problem for LAST_ACK state
951 * as ACK is intrepretated as ACKing last FIN.
953 if (old_state == TCP_CONNTRACK_LAST_ACK)
954 ct->proto.tcp.last_flags |=
955 IP_CT_EXP_CHALLENGE_ACK;
957 spin_unlock_bh(&ct->lock);
958 if (LOG_INVALID(net, IPPROTO_TCP))
959 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
960 "nf_ct_tcp: invalid packet ignored in "
961 "state %s ", tcp_conntrack_names[old_state]);
963 case TCP_CONNTRACK_MAX:
964 /* Special case for SYN proxy: when the SYN to the server or
965 * the SYN/ACK from the server is lost, the client may transmit
966 * a keep-alive packet while in SYN_SENT state. This needs to
967 * be associated with the original conntrack entry in order to
968 * generate a new SYN with the correct sequence number.
970 if (nfct_synproxy(ct) && old_state == TCP_CONNTRACK_SYN_SENT &&
971 index == TCP_ACK_SET && dir == IP_CT_DIR_ORIGINAL &&
972 ct->proto.tcp.last_dir == IP_CT_DIR_ORIGINAL &&
973 ct->proto.tcp.seen[dir].td_end - 1 == ntohl(th->seq)) {
974 pr_debug("nf_ct_tcp: SYN proxy client keep alive\n");
975 spin_unlock_bh(&ct->lock);
980 pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
981 dir, get_conntrack_index(th), old_state);
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: invalid state ");
987 case TCP_CONNTRACK_TIME_WAIT:
988 /* RFC5961 compliance cause stack to send "challenge-ACK"
989 * e.g. in response to spurious SYNs. Conntrack MUST
990 * not believe this ACK is acking last FIN.
992 if (old_state == TCP_CONNTRACK_LAST_ACK &&
993 index == TCP_ACK_SET &&
994 ct->proto.tcp.last_dir != dir &&
995 ct->proto.tcp.last_index == TCP_SYN_SET &&
996 (ct->proto.tcp.last_flags & IP_CT_EXP_CHALLENGE_ACK)) {
997 /* Detected RFC5961 challenge ACK */
998 ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK;
999 spin_unlock_bh(&ct->lock);
1000 if (LOG_INVALID(net, IPPROTO_TCP))
1001 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
1002 "nf_ct_tcp: challenge-ACK ignored ");
1003 return NF_ACCEPT; /* Don't change state */
1006 case TCP_CONNTRACK_CLOSE:
1007 if (index == TCP_RST_SET
1008 && (ct->proto.tcp.seen[!dir].flags & IP_CT_TCP_FLAG_MAXACK_SET)
1009 && before(ntohl(th->seq), ct->proto.tcp.seen[!dir].td_maxack)) {
1011 spin_unlock_bh(&ct->lock);
1012 if (LOG_INVALID(net, IPPROTO_TCP))
1013 nf_log_packet(net, pf, 0, skb, NULL, NULL,
1014 NULL, "nf_ct_tcp: invalid RST ");
1017 if (index == TCP_RST_SET
1018 && ((test_bit(IPS_SEEN_REPLY_BIT, &ct->status)
1019 && ct->proto.tcp.last_index == TCP_SYN_SET)
1020 || (!test_bit(IPS_ASSURED_BIT, &ct->status)
1021 && ct->proto.tcp.last_index == TCP_ACK_SET))
1022 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
1023 /* RST sent to invalid SYN or ACK we had let through
1024 * at a) and c) above:
1026 * a) SYN was in window then
1027 * c) we hold a half-open connection.
1029 * Delete our connection entry.
1030 * We skip window checking, because packet might ACK
1031 * segments we ignored. */
1034 /* Just fall through */
1036 /* Keep compilers happy. */
1040 if (!tcp_in_window(ct, &ct->proto.tcp, dir, index,
1041 skb, dataoff, th, pf)) {
1042 spin_unlock_bh(&ct->lock);
1046 /* From now on we have got in-window packets */
1047 ct->proto.tcp.last_index = index;
1048 ct->proto.tcp.last_dir = dir;
1050 pr_debug("tcp_conntracks: ");
1051 nf_ct_dump_tuple(tuple);
1052 pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
1053 (th->syn ? 1 : 0), (th->ack ? 1 : 0),
1054 (th->fin ? 1 : 0), (th->rst ? 1 : 0),
1055 old_state, new_state);
1057 ct->proto.tcp.state = new_state;
1058 if (old_state != new_state
1059 && new_state == TCP_CONNTRACK_FIN_WAIT)
1060 ct->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT;
1062 if (ct->proto.tcp.retrans >= tn->tcp_max_retrans &&
1063 timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS])
1064 timeout = timeouts[TCP_CONNTRACK_RETRANS];
1065 else if ((ct->proto.tcp.seen[0].flags | ct->proto.tcp.seen[1].flags) &
1066 IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED &&
1067 timeouts[new_state] > timeouts[TCP_CONNTRACK_UNACK])
1068 timeout = timeouts[TCP_CONNTRACK_UNACK];
1070 timeout = timeouts[new_state];
1071 spin_unlock_bh(&ct->lock);
1073 if (new_state != old_state)
1074 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
1076 if (!test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
1077 /* If only reply is a RST, we can consider ourselves not to
1078 have an established connection: this is a fairly common
1079 problem case, so we can delete the conntrack
1080 immediately. --RR */
1082 nf_ct_kill_acct(ct, ctinfo, skb);
1085 /* ESTABLISHED without SEEN_REPLY, i.e. mid-connection
1086 * pickup with loose=1. Avoid large ESTABLISHED timeout.
1088 if (new_state == TCP_CONNTRACK_ESTABLISHED &&
1089 timeout > timeouts[TCP_CONNTRACK_UNACK])
1090 timeout = timeouts[TCP_CONNTRACK_UNACK];
1091 } else if (!test_bit(IPS_ASSURED_BIT, &ct->status)
1092 && (old_state == TCP_CONNTRACK_SYN_RECV
1093 || old_state == TCP_CONNTRACK_ESTABLISHED)
1094 && new_state == TCP_CONNTRACK_ESTABLISHED) {
1095 /* Set ASSURED if we see see valid ack in ESTABLISHED
1096 after SYN_RECV or a valid answer for a picked up
1098 set_bit(IPS_ASSURED_BIT, &ct->status);
1099 nf_conntrack_event_cache(IPCT_ASSURED, ct);
1101 nf_ct_refresh_acct(ct, ctinfo, skb, timeout);
1106 /* Called when a new connection for this protocol found. */
1107 static bool tcp_new(struct nf_conn *ct, const struct sk_buff *skb,
1108 unsigned int dataoff, unsigned int *timeouts)
1110 enum tcp_conntrack new_state;
1111 const struct tcphdr *th;
1112 struct tcphdr _tcph;
1113 struct net *net = nf_ct_net(ct);
1114 struct nf_tcp_net *tn = tcp_pernet(net);
1115 const struct ip_ct_tcp_state *sender = &ct->proto.tcp.seen[0];
1116 const struct ip_ct_tcp_state *receiver = &ct->proto.tcp.seen[1];
1118 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
1121 /* Don't need lock here: this conntrack not in circulation yet */
1122 new_state = tcp_conntracks[0][get_conntrack_index(th)][TCP_CONNTRACK_NONE];
1124 /* Invalid: delete conntrack */
1125 if (new_state >= TCP_CONNTRACK_MAX) {
1126 pr_debug("nf_ct_tcp: invalid new deleting.\n");
1130 if (new_state == TCP_CONNTRACK_SYN_SENT) {
1131 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
1133 ct->proto.tcp.seen[0].td_end =
1134 segment_seq_plus_len(ntohl(th->seq), skb->len,
1136 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1137 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1138 ct->proto.tcp.seen[0].td_maxwin = 1;
1139 ct->proto.tcp.seen[0].td_maxend =
1140 ct->proto.tcp.seen[0].td_end;
1142 tcp_options(skb, dataoff, th, &ct->proto.tcp.seen[0]);
1143 } else if (tn->tcp_loose == 0) {
1144 /* Don't try to pick up connections. */
1147 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
1149 * We are in the middle of a connection,
1150 * its history is lost for us.
1151 * Let's try to use the data from the packet.
1153 ct->proto.tcp.seen[0].td_end =
1154 segment_seq_plus_len(ntohl(th->seq), skb->len,
1156 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1157 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1158 ct->proto.tcp.seen[0].td_maxwin = 1;
1159 ct->proto.tcp.seen[0].td_maxend =
1160 ct->proto.tcp.seen[0].td_end +
1161 ct->proto.tcp.seen[0].td_maxwin;
1163 /* We assume SACK and liberal window checking to handle
1165 ct->proto.tcp.seen[0].flags =
1166 ct->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM |
1167 IP_CT_TCP_FLAG_BE_LIBERAL;
1170 /* tcp_packet will set them */
1171 ct->proto.tcp.last_index = TCP_NONE_SET;
1173 pr_debug("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i "
1174 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
1175 sender->td_end, sender->td_maxend, sender->td_maxwin,
1177 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
1178 receiver->td_scale);
1182 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1184 #include <linux/netfilter/nfnetlink.h>
1185 #include <linux/netfilter/nfnetlink_conntrack.h>
1187 static int tcp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
1190 struct nlattr *nest_parms;
1191 struct nf_ct_tcp_flags tmp = {};
1193 spin_lock_bh(&ct->lock);
1194 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_TCP | NLA_F_NESTED);
1196 goto nla_put_failure;
1198 if (nla_put_u8(skb, CTA_PROTOINFO_TCP_STATE, ct->proto.tcp.state) ||
1199 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL,
1200 ct->proto.tcp.seen[0].td_scale) ||
1201 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY,
1202 ct->proto.tcp.seen[1].td_scale))
1203 goto nla_put_failure;
1205 tmp.flags = ct->proto.tcp.seen[0].flags;
1206 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL,
1207 sizeof(struct nf_ct_tcp_flags), &tmp))
1208 goto nla_put_failure;
1210 tmp.flags = ct->proto.tcp.seen[1].flags;
1211 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY,
1212 sizeof(struct nf_ct_tcp_flags), &tmp))
1213 goto nla_put_failure;
1214 spin_unlock_bh(&ct->lock);
1216 nla_nest_end(skb, nest_parms);
1221 spin_unlock_bh(&ct->lock);
1225 static const struct nla_policy tcp_nla_policy[CTA_PROTOINFO_TCP_MAX+1] = {
1226 [CTA_PROTOINFO_TCP_STATE] = { .type = NLA_U8 },
1227 [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] = { .type = NLA_U8 },
1228 [CTA_PROTOINFO_TCP_WSCALE_REPLY] = { .type = NLA_U8 },
1229 [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL] = { .len = sizeof(struct nf_ct_tcp_flags) },
1230 [CTA_PROTOINFO_TCP_FLAGS_REPLY] = { .len = sizeof(struct nf_ct_tcp_flags) },
1233 static int nlattr_to_tcp(struct nlattr *cda[], struct nf_conn *ct)
1235 struct nlattr *pattr = cda[CTA_PROTOINFO_TCP];
1236 struct nlattr *tb[CTA_PROTOINFO_TCP_MAX+1];
1239 /* updates could not contain anything about the private
1240 * protocol info, in that case skip the parsing */
1244 err = nla_parse_nested(tb, CTA_PROTOINFO_TCP_MAX, pattr, tcp_nla_policy);
1248 if (tb[CTA_PROTOINFO_TCP_STATE] &&
1249 nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]) >= TCP_CONNTRACK_MAX)
1252 spin_lock_bh(&ct->lock);
1253 if (tb[CTA_PROTOINFO_TCP_STATE])
1254 ct->proto.tcp.state = nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]);
1256 if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]) {
1257 struct nf_ct_tcp_flags *attr =
1258 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]);
1259 ct->proto.tcp.seen[0].flags &= ~attr->mask;
1260 ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask;
1263 if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]) {
1264 struct nf_ct_tcp_flags *attr =
1265 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]);
1266 ct->proto.tcp.seen[1].flags &= ~attr->mask;
1267 ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask;
1270 if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] &&
1271 tb[CTA_PROTOINFO_TCP_WSCALE_REPLY] &&
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) {
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]);
1279 spin_unlock_bh(&ct->lock);
1284 static int tcp_nlattr_size(void)
1286 return nla_total_size(0) /* CTA_PROTOINFO_TCP */
1287 + nla_policy_len(tcp_nla_policy, CTA_PROTOINFO_TCP_MAX + 1);
1290 static int tcp_nlattr_tuple_size(void)
1292 return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1);
1296 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1298 #include <linux/netfilter/nfnetlink.h>
1299 #include <linux/netfilter/nfnetlink_cttimeout.h>
1301 static int tcp_timeout_nlattr_to_obj(struct nlattr *tb[],
1302 struct net *net, void *data)
1304 unsigned int *timeouts = data;
1305 struct nf_tcp_net *tn = tcp_pernet(net);
1308 /* set default TCP timeouts. */
1309 for (i=0; i<TCP_CONNTRACK_TIMEOUT_MAX; i++)
1310 timeouts[i] = tn->timeouts[i];
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;
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;
1320 if (tb[CTA_TIMEOUT_TCP_ESTABLISHED]) {
1321 timeouts[TCP_CONNTRACK_ESTABLISHED] =
1322 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_ESTABLISHED]))*HZ;
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;
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;
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;
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;
1340 if (tb[CTA_TIMEOUT_TCP_CLOSE]) {
1341 timeouts[TCP_CONNTRACK_CLOSE] =
1342 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE]))*HZ;
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;
1348 if (tb[CTA_TIMEOUT_TCP_RETRANS]) {
1349 timeouts[TCP_CONNTRACK_RETRANS] =
1350 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_RETRANS]))*HZ;
1352 if (tb[CTA_TIMEOUT_TCP_UNACK]) {
1353 timeouts[TCP_CONNTRACK_UNACK] =
1354 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_UNACK]))*HZ;
1360 tcp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
1362 const unsigned int *timeouts = data;
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;
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 },
1403 [CTA_TIMEOUT_TCP_RETRANS] = { .type = NLA_U32 },
1404 [CTA_TIMEOUT_TCP_UNACK] = { .type = NLA_U32 },
1406 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1408 #ifdef CONFIG_SYSCTL
1409 static struct ctl_table tcp_sysctl_table[] = {
1411 .procname = "nf_conntrack_tcp_timeout_syn_sent",
1412 .maxlen = sizeof(unsigned int),
1414 .proc_handler = proc_dointvec_jiffies,
1417 .procname = "nf_conntrack_tcp_timeout_syn_recv",
1418 .maxlen = sizeof(unsigned int),
1420 .proc_handler = proc_dointvec_jiffies,
1423 .procname = "nf_conntrack_tcp_timeout_established",
1424 .maxlen = sizeof(unsigned int),
1426 .proc_handler = proc_dointvec_jiffies,
1429 .procname = "nf_conntrack_tcp_timeout_fin_wait",
1430 .maxlen = sizeof(unsigned int),
1432 .proc_handler = proc_dointvec_jiffies,
1435 .procname = "nf_conntrack_tcp_timeout_close_wait",
1436 .maxlen = sizeof(unsigned int),
1438 .proc_handler = proc_dointvec_jiffies,
1441 .procname = "nf_conntrack_tcp_timeout_last_ack",
1442 .maxlen = sizeof(unsigned int),
1444 .proc_handler = proc_dointvec_jiffies,
1447 .procname = "nf_conntrack_tcp_timeout_time_wait",
1448 .maxlen = sizeof(unsigned int),
1450 .proc_handler = proc_dointvec_jiffies,
1453 .procname = "nf_conntrack_tcp_timeout_close",
1454 .maxlen = sizeof(unsigned int),
1456 .proc_handler = proc_dointvec_jiffies,
1459 .procname = "nf_conntrack_tcp_timeout_max_retrans",
1460 .maxlen = sizeof(unsigned int),
1462 .proc_handler = proc_dointvec_jiffies,
1465 .procname = "nf_conntrack_tcp_timeout_unacknowledged",
1466 .maxlen = sizeof(unsigned int),
1468 .proc_handler = proc_dointvec_jiffies,
1471 .procname = "nf_conntrack_tcp_loose",
1472 .maxlen = sizeof(unsigned int),
1474 .proc_handler = proc_dointvec,
1477 .procname = "nf_conntrack_tcp_be_liberal",
1478 .maxlen = sizeof(unsigned int),
1480 .proc_handler = proc_dointvec,
1483 .procname = "nf_conntrack_tcp_max_retrans",
1484 .maxlen = sizeof(unsigned int),
1486 .proc_handler = proc_dointvec,
1491 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
1492 static struct ctl_table tcp_compat_sysctl_table[] = {
1494 .procname = "ip_conntrack_tcp_timeout_syn_sent",
1495 .maxlen = sizeof(unsigned int),
1497 .proc_handler = proc_dointvec_jiffies,
1500 .procname = "ip_conntrack_tcp_timeout_syn_sent2",
1501 .maxlen = sizeof(unsigned int),
1503 .proc_handler = proc_dointvec_jiffies,
1506 .procname = "ip_conntrack_tcp_timeout_syn_recv",
1507 .maxlen = sizeof(unsigned int),
1509 .proc_handler = proc_dointvec_jiffies,
1512 .procname = "ip_conntrack_tcp_timeout_established",
1513 .maxlen = sizeof(unsigned int),
1515 .proc_handler = proc_dointvec_jiffies,
1518 .procname = "ip_conntrack_tcp_timeout_fin_wait",
1519 .maxlen = sizeof(unsigned int),
1521 .proc_handler = proc_dointvec_jiffies,
1524 .procname = "ip_conntrack_tcp_timeout_close_wait",
1525 .maxlen = sizeof(unsigned int),
1527 .proc_handler = proc_dointvec_jiffies,
1530 .procname = "ip_conntrack_tcp_timeout_last_ack",
1531 .maxlen = sizeof(unsigned int),
1533 .proc_handler = proc_dointvec_jiffies,
1536 .procname = "ip_conntrack_tcp_timeout_time_wait",
1537 .maxlen = sizeof(unsigned int),
1539 .proc_handler = proc_dointvec_jiffies,
1542 .procname = "ip_conntrack_tcp_timeout_close",
1543 .maxlen = sizeof(unsigned int),
1545 .proc_handler = proc_dointvec_jiffies,
1548 .procname = "ip_conntrack_tcp_timeout_max_retrans",
1549 .maxlen = sizeof(unsigned int),
1551 .proc_handler = proc_dointvec_jiffies,
1554 .procname = "ip_conntrack_tcp_loose",
1555 .maxlen = sizeof(unsigned int),
1557 .proc_handler = proc_dointvec,
1560 .procname = "ip_conntrack_tcp_be_liberal",
1561 .maxlen = sizeof(unsigned int),
1563 .proc_handler = proc_dointvec,
1566 .procname = "ip_conntrack_tcp_max_retrans",
1567 .maxlen = sizeof(unsigned int),
1569 .proc_handler = proc_dointvec,
1573 #endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
1574 #endif /* CONFIG_SYSCTL */
1576 static int tcp_kmemdup_sysctl_table(struct nf_proto_net *pn,
1577 struct nf_tcp_net *tn)
1579 #ifdef CONFIG_SYSCTL
1583 pn->ctl_table = kmemdup(tcp_sysctl_table,
1584 sizeof(tcp_sysctl_table),
1589 pn->ctl_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1590 pn->ctl_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1591 pn->ctl_table[2].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1592 pn->ctl_table[3].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1593 pn->ctl_table[4].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1594 pn->ctl_table[5].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1595 pn->ctl_table[6].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1596 pn->ctl_table[7].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1597 pn->ctl_table[8].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1598 pn->ctl_table[9].data = &tn->timeouts[TCP_CONNTRACK_UNACK];
1599 pn->ctl_table[10].data = &tn->tcp_loose;
1600 pn->ctl_table[11].data = &tn->tcp_be_liberal;
1601 pn->ctl_table[12].data = &tn->tcp_max_retrans;
1606 static int tcp_kmemdup_compat_sysctl_table(struct nf_proto_net *pn,
1607 struct nf_tcp_net *tn)
1609 #ifdef CONFIG_SYSCTL
1610 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
1611 pn->ctl_compat_table = kmemdup(tcp_compat_sysctl_table,
1612 sizeof(tcp_compat_sysctl_table),
1614 if (!pn->ctl_compat_table)
1617 pn->ctl_compat_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1618 pn->ctl_compat_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT2];
1619 pn->ctl_compat_table[2].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1620 pn->ctl_compat_table[3].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1621 pn->ctl_compat_table[4].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1622 pn->ctl_compat_table[5].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1623 pn->ctl_compat_table[6].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1624 pn->ctl_compat_table[7].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1625 pn->ctl_compat_table[8].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1626 pn->ctl_compat_table[9].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1627 pn->ctl_compat_table[10].data = &tn->tcp_loose;
1628 pn->ctl_compat_table[11].data = &tn->tcp_be_liberal;
1629 pn->ctl_compat_table[12].data = &tn->tcp_max_retrans;
1635 static int tcp_init_net(struct net *net, u_int16_t proto)
1638 struct nf_tcp_net *tn = tcp_pernet(net);
1639 struct nf_proto_net *pn = &tn->pn;
1644 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
1645 tn->timeouts[i] = tcp_timeouts[i];
1647 tn->tcp_loose = nf_ct_tcp_loose;
1648 tn->tcp_be_liberal = nf_ct_tcp_be_liberal;
1649 tn->tcp_max_retrans = nf_ct_tcp_max_retrans;
1652 if (proto == AF_INET) {
1653 ret = tcp_kmemdup_compat_sysctl_table(pn, tn);
1657 ret = tcp_kmemdup_sysctl_table(pn, tn);
1659 nf_ct_kfree_compat_sysctl_table(pn);
1661 ret = tcp_kmemdup_sysctl_table(pn, tn);
1666 static struct nf_proto_net *tcp_get_net_proto(struct net *net)
1668 return &net->ct.nf_ct_proto.tcp.pn;
1671 struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp4 __read_mostly =
1674 .l4proto = IPPROTO_TCP,
1676 .pkt_to_tuple = tcp_pkt_to_tuple,
1677 .invert_tuple = tcp_invert_tuple,
1678 .print_tuple = tcp_print_tuple,
1679 .print_conntrack = tcp_print_conntrack,
1680 .packet = tcp_packet,
1681 .get_timeouts = tcp_get_timeouts,
1684 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1685 .to_nlattr = tcp_to_nlattr,
1686 .nlattr_size = tcp_nlattr_size,
1687 .from_nlattr = nlattr_to_tcp,
1688 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1689 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
1690 .nlattr_tuple_size = tcp_nlattr_tuple_size,
1691 .nla_policy = nf_ct_port_nla_policy,
1693 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1695 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1696 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1697 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1698 .obj_size = sizeof(unsigned int) *
1699 TCP_CONNTRACK_TIMEOUT_MAX,
1700 .nla_policy = tcp_timeout_nla_policy,
1702 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1703 .init_net = tcp_init_net,
1704 .get_net_proto = tcp_get_net_proto,
1706 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp4);
1708 struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6 __read_mostly =
1710 .l3proto = PF_INET6,
1711 .l4proto = IPPROTO_TCP,
1713 .pkt_to_tuple = tcp_pkt_to_tuple,
1714 .invert_tuple = tcp_invert_tuple,
1715 .print_tuple = tcp_print_tuple,
1716 .print_conntrack = tcp_print_conntrack,
1717 .packet = tcp_packet,
1718 .get_timeouts = tcp_get_timeouts,
1721 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1722 .to_nlattr = tcp_to_nlattr,
1723 .nlattr_size = tcp_nlattr_size,
1724 .from_nlattr = nlattr_to_tcp,
1725 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1726 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
1727 .nlattr_tuple_size = tcp_nlattr_tuple_size,
1728 .nla_policy = nf_ct_port_nla_policy,
1730 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1732 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1733 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1734 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1735 .obj_size = sizeof(unsigned int) *
1736 TCP_CONNTRACK_TIMEOUT_MAX,
1737 .nla_policy = tcp_timeout_nla_policy,
1739 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1740 .init_net = tcp_init_net,
1741 .get_net_proto = tcp_get_net_proto,
1743 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp6);