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),
387 state->flags &= IP_CT_TCP_FLAG_BE_LIBERAL;
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,
499 struct net *net = nf_ct_net(ct);
500 struct nf_tcp_net *tn = tcp_pernet(net);
501 struct ip_ct_tcp_state *sender = &state->seen[dir];
502 struct ip_ct_tcp_state *receiver = &state->seen[!dir];
503 const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
504 __u32 seq, ack, sack, end, win, swin;
506 bool res, in_recv_win;
509 * Get the required data from the packet.
511 seq = ntohl(tcph->seq);
512 ack = sack = ntohl(tcph->ack_seq);
513 win = ntohs(tcph->window);
514 end = segment_seq_plus_len(seq, skb->len, dataoff, tcph);
516 if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM)
517 tcp_sack(skb, dataoff, tcph, &sack);
519 /* Take into account NAT sequence number mangling */
520 receiver_offset = nf_ct_seq_offset(ct, !dir, ack - 1);
521 ack -= receiver_offset;
522 sack -= receiver_offset;
524 pr_debug("tcp_in_window: START\n");
525 pr_debug("tcp_in_window: ");
526 nf_ct_dump_tuple(tuple);
527 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
528 seq, ack, receiver_offset, sack, receiver_offset, win, end);
529 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
530 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
531 sender->td_end, sender->td_maxend, sender->td_maxwin,
533 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
536 if (sender->td_maxwin == 0) {
538 * Initialize sender data.
542 * SYN-ACK in reply to a SYN
543 * or SYN from reply direction in simultaneous open.
546 sender->td_maxend = end;
547 sender->td_maxwin = (win == 0 ? 1 : win);
549 tcp_options(skb, dataoff, tcph, sender);
552 * Both sides must send the Window Scale option
553 * to enable window scaling in either direction.
555 if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE
556 && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE))
558 receiver->td_scale = 0;
560 /* Simultaneous open */
564 * We are in the middle of a connection,
565 * its history is lost for us.
566 * Let's try to use the data from the packet.
568 sender->td_end = end;
569 swin = win << sender->td_scale;
570 sender->td_maxwin = (swin == 0 ? 1 : swin);
571 sender->td_maxend = end + sender->td_maxwin;
573 * We haven't seen traffic in the other direction yet
574 * but we have to tweak window tracking to pass III
575 * and IV until that happens.
577 if (receiver->td_maxwin == 0)
578 receiver->td_end = receiver->td_maxend = sack;
580 } else if (((state->state == TCP_CONNTRACK_SYN_SENT
581 && dir == IP_CT_DIR_ORIGINAL)
582 || (state->state == TCP_CONNTRACK_SYN_RECV
583 && dir == IP_CT_DIR_REPLY))
584 && after(end, sender->td_end)) {
586 * RFC 793: "if a TCP is reinitialized ... then it need
587 * not wait at all; it must only be sure to use sequence
588 * numbers larger than those recently used."
591 sender->td_maxend = end;
592 sender->td_maxwin = (win == 0 ? 1 : win);
594 tcp_options(skb, dataoff, tcph, sender);
599 * If there is no ACK, just pretend it was set and OK.
601 ack = sack = receiver->td_end;
602 } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) ==
603 (TCP_FLAG_ACK|TCP_FLAG_RST))
606 * Broken TCP stacks, that set ACK in RST packets as well
607 * with zero ack value.
609 ack = sack = receiver->td_end;
612 if (tcph->rst && seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT)
614 * RST sent answering SYN.
616 seq = end = sender->td_end;
618 pr_debug("tcp_in_window: ");
619 nf_ct_dump_tuple(tuple);
620 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
621 seq, ack, receiver_offset, sack, receiver_offset, win, end);
622 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
623 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
624 sender->td_end, sender->td_maxend, sender->td_maxwin,
626 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
629 /* Is the ending sequence in the receive window (if available)? */
630 in_recv_win = !receiver->td_maxwin ||
631 after(end, sender->td_end - receiver->td_maxwin - 1);
633 pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
634 before(seq, sender->td_maxend + 1),
635 (in_recv_win ? 1 : 0),
636 before(sack, receiver->td_end + 1),
637 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1));
639 if (before(seq, sender->td_maxend + 1) &&
641 before(sack, receiver->td_end + 1) &&
642 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)) {
644 * Take into account window scaling (RFC 1323).
647 win <<= sender->td_scale;
650 * Update sender data.
652 swin = win + (sack - ack);
653 if (sender->td_maxwin < swin)
654 sender->td_maxwin = swin;
655 if (after(end, sender->td_end)) {
656 sender->td_end = end;
657 sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
660 if (!(sender->flags & IP_CT_TCP_FLAG_MAXACK_SET)) {
661 sender->td_maxack = ack;
662 sender->flags |= IP_CT_TCP_FLAG_MAXACK_SET;
663 } else if (after(ack, sender->td_maxack))
664 sender->td_maxack = ack;
668 * Update receiver data.
670 if (receiver->td_maxwin != 0 && after(end, sender->td_maxend))
671 receiver->td_maxwin += end - sender->td_maxend;
672 if (after(sack + win, receiver->td_maxend - 1)) {
673 receiver->td_maxend = sack + win;
675 receiver->td_maxend++;
677 if (ack == receiver->td_end)
678 receiver->flags &= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
681 * Check retransmissions.
683 if (index == TCP_ACK_SET) {
684 if (state->last_dir == dir
685 && state->last_seq == seq
686 && state->last_ack == ack
687 && state->last_end == end
688 && state->last_win == win)
691 state->last_dir = dir;
692 state->last_seq = seq;
693 state->last_ack = ack;
694 state->last_end = end;
695 state->last_win = win;
702 if (sender->flags & IP_CT_TCP_FLAG_BE_LIBERAL ||
705 if (!res && LOG_INVALID(net, IPPROTO_TCP))
706 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
708 before(seq, sender->td_maxend + 1) ?
710 before(sack, receiver->td_end + 1) ?
711 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1) ? "BUG"
712 : "ACK is under the lower bound (possible overly delayed ACK)"
713 : "ACK is over the upper bound (ACKed data not seen yet)"
714 : "SEQ is under the lower bound (already ACKed data retransmitted)"
715 : "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 /* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
742 static int tcp_error(struct net *net, struct nf_conn *tmpl,
744 unsigned int dataoff,
746 unsigned int hooknum)
748 const struct tcphdr *th;
750 unsigned int tcplen = skb->len - dataoff;
753 /* Smaller that minimal TCP header? */
754 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
756 if (LOG_INVALID(net, IPPROTO_TCP))
757 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
758 "nf_ct_tcp: short packet ");
762 /* Not whole TCP header or malformed packet */
763 if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) {
764 if (LOG_INVALID(net, IPPROTO_TCP))
765 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
766 "nf_ct_tcp: truncated/malformed packet ");
770 /* Checksum invalid? Ignore.
771 * We skip checking packets on the outgoing path
772 * because the checksum is assumed to be correct.
774 /* FIXME: Source route IP option packets --RR */
775 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
776 nf_checksum(skb, hooknum, dataoff, IPPROTO_TCP, pf)) {
777 if (LOG_INVALID(net, IPPROTO_TCP))
778 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
779 "nf_ct_tcp: bad TCP checksum ");
783 /* Check TCP flags. */
784 tcpflags = (tcp_flag_byte(th) & ~(TCPHDR_ECE|TCPHDR_CWR|TCPHDR_PSH));
785 if (!tcp_valid_flags[tcpflags]) {
786 if (LOG_INVALID(net, IPPROTO_TCP))
787 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
788 "nf_ct_tcp: invalid TCP flag combination ");
795 static unsigned int *tcp_get_timeouts(struct net *net)
797 return tcp_pernet(net)->timeouts;
800 static void nf_ct_tcp_state_reset(struct ip_ct_tcp_state *state)
803 state->td_maxend = 0;
804 state->td_maxwin = 0;
805 state->td_maxack = 0;
807 state->flags &= IP_CT_TCP_FLAG_BE_LIBERAL;
810 /* Returns verdict for packet, or -1 for invalid. */
811 static int tcp_packet(struct nf_conn *ct,
812 const struct sk_buff *skb,
813 unsigned int dataoff,
814 enum ip_conntrack_info ctinfo,
816 unsigned int *timeouts)
818 struct net *net = nf_ct_net(ct);
819 struct nf_tcp_net *tn = tcp_pernet(net);
820 struct nf_conntrack_tuple *tuple;
821 enum tcp_conntrack new_state, old_state;
822 enum ip_conntrack_dir dir;
823 const struct tcphdr *th;
825 unsigned long timeout;
828 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
831 spin_lock_bh(&ct->lock);
832 old_state = ct->proto.tcp.state;
833 dir = CTINFO2DIR(ctinfo);
834 index = get_conntrack_index(th);
835 new_state = tcp_conntracks[dir][index][old_state];
836 tuple = &ct->tuplehash[dir].tuple;
839 case TCP_CONNTRACK_SYN_SENT:
840 if (old_state < TCP_CONNTRACK_TIME_WAIT)
842 /* RFC 1122: "When a connection is closed actively,
843 * it MUST linger in TIME-WAIT state for a time 2xMSL
844 * (Maximum Segment Lifetime). However, it MAY accept
845 * a new SYN from the remote TCP to reopen the connection
846 * directly from TIME-WAIT state, if..."
847 * We ignore the conditions because we are in the
848 * TIME-WAIT state anyway.
850 * Handle aborted connections: we and the server
851 * think there is an existing connection but the client
852 * aborts it and starts a new one.
854 if (((ct->proto.tcp.seen[dir].flags
855 | ct->proto.tcp.seen[!dir].flags)
856 & IP_CT_TCP_FLAG_CLOSE_INIT)
857 || (ct->proto.tcp.last_dir == dir
858 && ct->proto.tcp.last_index == TCP_RST_SET)) {
859 /* Attempt to reopen a closed/aborted connection.
860 * Delete this connection and look up again. */
861 spin_unlock_bh(&ct->lock);
863 /* Only repeat if we can actually remove the timer.
864 * Destruction may already be in progress in process
865 * context and we must give it a chance to terminate.
872 case TCP_CONNTRACK_IGNORE:
875 * Our connection entry may be out of sync, so ignore
876 * packets which may signal the real connection between
877 * the client and the server.
880 * b) SYN/ACK in REPLY
881 * c) ACK in reply direction after initial SYN in original.
883 * If the ignored packet is invalid, the receiver will send
884 * a RST we'll catch below.
886 if (index == TCP_SYNACK_SET
887 && ct->proto.tcp.last_index == TCP_SYN_SET
888 && ct->proto.tcp.last_dir != dir
889 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
890 /* b) This SYN/ACK acknowledges a SYN that we earlier
891 * ignored as invalid. This means that the client and
892 * the server are both in sync, while the firewall is
893 * not. We get in sync from the previously annotated
896 old_state = TCP_CONNTRACK_SYN_SENT;
897 new_state = TCP_CONNTRACK_SYN_RECV;
898 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_end =
899 ct->proto.tcp.last_end;
900 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxend =
901 ct->proto.tcp.last_end;
902 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxwin =
903 ct->proto.tcp.last_win == 0 ?
904 1 : ct->proto.tcp.last_win;
905 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_scale =
906 ct->proto.tcp.last_wscale;
907 ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK;
908 ct->proto.tcp.seen[ct->proto.tcp.last_dir].flags =
909 ct->proto.tcp.last_flags;
910 nf_ct_tcp_state_reset(&ct->proto.tcp.seen[dir]);
913 ct->proto.tcp.last_index = index;
914 ct->proto.tcp.last_dir = dir;
915 ct->proto.tcp.last_seq = ntohl(th->seq);
916 ct->proto.tcp.last_end =
917 segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th);
918 ct->proto.tcp.last_win = ntohs(th->window);
920 /* a) This is a SYN in ORIGINAL. The client and the server
921 * may be in sync but we are not. In that case, we annotate
922 * the TCP options and let the packet go through. If it is a
923 * valid SYN packet, the server will reply with a SYN/ACK, and
924 * then we'll get in sync. Otherwise, the server potentially
925 * responds with a challenge ACK if implementing RFC5961.
927 if (index == TCP_SYN_SET && dir == IP_CT_DIR_ORIGINAL) {
928 struct ip_ct_tcp_state seen = {};
930 ct->proto.tcp.last_flags =
931 ct->proto.tcp.last_wscale = 0;
932 tcp_options(skb, dataoff, th, &seen);
933 if (seen.flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
934 ct->proto.tcp.last_flags |=
935 IP_CT_TCP_FLAG_WINDOW_SCALE;
936 ct->proto.tcp.last_wscale = seen.td_scale;
938 if (seen.flags & IP_CT_TCP_FLAG_SACK_PERM) {
939 ct->proto.tcp.last_flags |=
940 IP_CT_TCP_FLAG_SACK_PERM;
942 /* Mark the potential for RFC5961 challenge ACK,
943 * this pose a special problem for LAST_ACK state
944 * as ACK is intrepretated as ACKing last FIN.
946 if (old_state == TCP_CONNTRACK_LAST_ACK)
947 ct->proto.tcp.last_flags |=
948 IP_CT_EXP_CHALLENGE_ACK;
950 spin_unlock_bh(&ct->lock);
951 if (LOG_INVALID(net, IPPROTO_TCP))
952 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
953 "nf_ct_tcp: invalid packet ignored in "
954 "state %s ", tcp_conntrack_names[old_state]);
956 case TCP_CONNTRACK_MAX:
957 /* Special case for SYN proxy: when the SYN to the server or
958 * the SYN/ACK from the server is lost, the client may transmit
959 * a keep-alive packet while in SYN_SENT state. This needs to
960 * be associated with the original conntrack entry in order to
961 * generate a new SYN with the correct sequence number.
963 if (nfct_synproxy(ct) && old_state == TCP_CONNTRACK_SYN_SENT &&
964 index == TCP_ACK_SET && dir == IP_CT_DIR_ORIGINAL &&
965 ct->proto.tcp.last_dir == IP_CT_DIR_ORIGINAL &&
966 ct->proto.tcp.seen[dir].td_end - 1 == ntohl(th->seq)) {
967 pr_debug("nf_ct_tcp: SYN proxy client keep alive\n");
968 spin_unlock_bh(&ct->lock);
973 pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
974 dir, get_conntrack_index(th), old_state);
975 spin_unlock_bh(&ct->lock);
976 if (LOG_INVALID(net, IPPROTO_TCP))
977 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
978 "nf_ct_tcp: invalid state ");
980 case TCP_CONNTRACK_TIME_WAIT:
981 /* RFC5961 compliance cause stack to send "challenge-ACK"
982 * e.g. in response to spurious SYNs. Conntrack MUST
983 * not believe this ACK is acking last FIN.
985 if (old_state == TCP_CONNTRACK_LAST_ACK &&
986 index == TCP_ACK_SET &&
987 ct->proto.tcp.last_dir != dir &&
988 ct->proto.tcp.last_index == TCP_SYN_SET &&
989 (ct->proto.tcp.last_flags & IP_CT_EXP_CHALLENGE_ACK)) {
990 /* Detected RFC5961 challenge ACK */
991 ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK;
992 spin_unlock_bh(&ct->lock);
993 if (LOG_INVALID(net, IPPROTO_TCP))
994 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
995 "nf_ct_tcp: challenge-ACK ignored ");
996 return NF_ACCEPT; /* Don't change state */
999 case TCP_CONNTRACK_CLOSE:
1000 if (index == TCP_RST_SET
1001 && (ct->proto.tcp.seen[!dir].flags & IP_CT_TCP_FLAG_MAXACK_SET)
1002 && before(ntohl(th->seq), ct->proto.tcp.seen[!dir].td_maxack)) {
1004 spin_unlock_bh(&ct->lock);
1005 if (LOG_INVALID(net, IPPROTO_TCP))
1006 nf_log_packet(net, pf, 0, skb, NULL, NULL,
1007 NULL, "nf_ct_tcp: invalid RST ");
1010 if (index == TCP_RST_SET
1011 && ((test_bit(IPS_SEEN_REPLY_BIT, &ct->status)
1012 && ct->proto.tcp.last_index == TCP_SYN_SET)
1013 || (!test_bit(IPS_ASSURED_BIT, &ct->status)
1014 && ct->proto.tcp.last_index == TCP_ACK_SET))
1015 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
1016 /* RST sent to invalid SYN or ACK we had let through
1017 * at a) and c) above:
1019 * a) SYN was in window then
1020 * c) we hold a half-open connection.
1022 * Delete our connection entry.
1023 * We skip window checking, because packet might ACK
1024 * segments we ignored. */
1027 /* Just fall through */
1029 /* Keep compilers happy. */
1033 if (!tcp_in_window(ct, &ct->proto.tcp, dir, index,
1034 skb, dataoff, th, pf)) {
1035 spin_unlock_bh(&ct->lock);
1039 /* From now on we have got in-window packets */
1040 ct->proto.tcp.last_index = index;
1041 ct->proto.tcp.last_dir = dir;
1043 pr_debug("tcp_conntracks: ");
1044 nf_ct_dump_tuple(tuple);
1045 pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
1046 (th->syn ? 1 : 0), (th->ack ? 1 : 0),
1047 (th->fin ? 1 : 0), (th->rst ? 1 : 0),
1048 old_state, new_state);
1050 ct->proto.tcp.state = new_state;
1051 if (old_state != new_state
1052 && new_state == TCP_CONNTRACK_FIN_WAIT)
1053 ct->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT;
1055 if (ct->proto.tcp.retrans >= tn->tcp_max_retrans &&
1056 timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS])
1057 timeout = timeouts[TCP_CONNTRACK_RETRANS];
1058 else if ((ct->proto.tcp.seen[0].flags | ct->proto.tcp.seen[1].flags) &
1059 IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED &&
1060 timeouts[new_state] > timeouts[TCP_CONNTRACK_UNACK])
1061 timeout = timeouts[TCP_CONNTRACK_UNACK];
1063 timeout = timeouts[new_state];
1064 spin_unlock_bh(&ct->lock);
1066 if (new_state != old_state)
1067 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
1069 if (!test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
1070 /* If only reply is a RST, we can consider ourselves not to
1071 have an established connection: this is a fairly common
1072 problem case, so we can delete the conntrack
1073 immediately. --RR */
1075 nf_ct_kill_acct(ct, ctinfo, skb);
1078 /* ESTABLISHED without SEEN_REPLY, i.e. mid-connection
1079 * pickup with loose=1. Avoid large ESTABLISHED timeout.
1081 if (new_state == TCP_CONNTRACK_ESTABLISHED &&
1082 timeout > timeouts[TCP_CONNTRACK_UNACK])
1083 timeout = timeouts[TCP_CONNTRACK_UNACK];
1084 } else if (!test_bit(IPS_ASSURED_BIT, &ct->status)
1085 && (old_state == TCP_CONNTRACK_SYN_RECV
1086 || old_state == TCP_CONNTRACK_ESTABLISHED)
1087 && new_state == TCP_CONNTRACK_ESTABLISHED) {
1088 /* Set ASSURED if we see see valid ack in ESTABLISHED
1089 after SYN_RECV or a valid answer for a picked up
1091 set_bit(IPS_ASSURED_BIT, &ct->status);
1092 nf_conntrack_event_cache(IPCT_ASSURED, ct);
1094 nf_ct_refresh_acct(ct, ctinfo, skb, timeout);
1099 /* Called when a new connection for this protocol found. */
1100 static bool tcp_new(struct nf_conn *ct, const struct sk_buff *skb,
1101 unsigned int dataoff, unsigned int *timeouts)
1103 enum tcp_conntrack new_state;
1104 const struct tcphdr *th;
1105 struct tcphdr _tcph;
1106 struct net *net = nf_ct_net(ct);
1107 struct nf_tcp_net *tn = tcp_pernet(net);
1108 const struct ip_ct_tcp_state *sender = &ct->proto.tcp.seen[0];
1109 const struct ip_ct_tcp_state *receiver = &ct->proto.tcp.seen[1];
1111 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
1114 /* Don't need lock here: this conntrack not in circulation yet */
1115 new_state = tcp_conntracks[0][get_conntrack_index(th)][TCP_CONNTRACK_NONE];
1117 /* Invalid: delete conntrack */
1118 if (new_state >= TCP_CONNTRACK_MAX) {
1119 pr_debug("nf_ct_tcp: invalid new deleting.\n");
1123 if (new_state == TCP_CONNTRACK_SYN_SENT) {
1124 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
1126 ct->proto.tcp.seen[0].td_end =
1127 segment_seq_plus_len(ntohl(th->seq), skb->len,
1129 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1130 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1131 ct->proto.tcp.seen[0].td_maxwin = 1;
1132 ct->proto.tcp.seen[0].td_maxend =
1133 ct->proto.tcp.seen[0].td_end;
1135 tcp_options(skb, dataoff, th, &ct->proto.tcp.seen[0]);
1136 } else if (tn->tcp_loose == 0) {
1137 /* Don't try to pick up connections. */
1140 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
1142 * We are in the middle of a connection,
1143 * its history is lost for us.
1144 * Let's try to use the data from the packet.
1146 ct->proto.tcp.seen[0].td_end =
1147 segment_seq_plus_len(ntohl(th->seq), skb->len,
1149 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1150 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1151 ct->proto.tcp.seen[0].td_maxwin = 1;
1152 ct->proto.tcp.seen[0].td_maxend =
1153 ct->proto.tcp.seen[0].td_end +
1154 ct->proto.tcp.seen[0].td_maxwin;
1156 /* We assume SACK and liberal window checking to handle
1158 ct->proto.tcp.seen[0].flags =
1159 ct->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM |
1160 IP_CT_TCP_FLAG_BE_LIBERAL;
1163 /* tcp_packet will set them */
1164 ct->proto.tcp.last_index = TCP_NONE_SET;
1166 pr_debug("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i "
1167 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
1168 sender->td_end, sender->td_maxend, sender->td_maxwin,
1170 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
1171 receiver->td_scale);
1175 static bool tcp_can_early_drop(const struct nf_conn *ct)
1177 switch (ct->proto.tcp.state) {
1178 case TCP_CONNTRACK_FIN_WAIT:
1179 case TCP_CONNTRACK_LAST_ACK:
1180 case TCP_CONNTRACK_TIME_WAIT:
1181 case TCP_CONNTRACK_CLOSE:
1182 case TCP_CONNTRACK_CLOSE_WAIT:
1191 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1193 #include <linux/netfilter/nfnetlink.h>
1194 #include <linux/netfilter/nfnetlink_conntrack.h>
1196 static int tcp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
1199 struct nlattr *nest_parms;
1200 struct nf_ct_tcp_flags tmp = {};
1202 spin_lock_bh(&ct->lock);
1203 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_TCP | NLA_F_NESTED);
1205 goto nla_put_failure;
1207 if (nla_put_u8(skb, CTA_PROTOINFO_TCP_STATE, ct->proto.tcp.state) ||
1208 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL,
1209 ct->proto.tcp.seen[0].td_scale) ||
1210 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY,
1211 ct->proto.tcp.seen[1].td_scale))
1212 goto nla_put_failure;
1214 tmp.flags = ct->proto.tcp.seen[0].flags;
1215 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL,
1216 sizeof(struct nf_ct_tcp_flags), &tmp))
1217 goto nla_put_failure;
1219 tmp.flags = ct->proto.tcp.seen[1].flags;
1220 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY,
1221 sizeof(struct nf_ct_tcp_flags), &tmp))
1222 goto nla_put_failure;
1223 spin_unlock_bh(&ct->lock);
1225 nla_nest_end(skb, nest_parms);
1230 spin_unlock_bh(&ct->lock);
1234 static const struct nla_policy tcp_nla_policy[CTA_PROTOINFO_TCP_MAX+1] = {
1235 [CTA_PROTOINFO_TCP_STATE] = { .type = NLA_U8 },
1236 [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] = { .type = NLA_U8 },
1237 [CTA_PROTOINFO_TCP_WSCALE_REPLY] = { .type = NLA_U8 },
1238 [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL] = { .len = sizeof(struct nf_ct_tcp_flags) },
1239 [CTA_PROTOINFO_TCP_FLAGS_REPLY] = { .len = sizeof(struct nf_ct_tcp_flags) },
1242 static int nlattr_to_tcp(struct nlattr *cda[], struct nf_conn *ct)
1244 struct nlattr *pattr = cda[CTA_PROTOINFO_TCP];
1245 struct nlattr *tb[CTA_PROTOINFO_TCP_MAX+1];
1248 /* updates could not contain anything about the private
1249 * protocol info, in that case skip the parsing */
1253 err = nla_parse_nested(tb, CTA_PROTOINFO_TCP_MAX, pattr,
1254 tcp_nla_policy, NULL);
1258 if (tb[CTA_PROTOINFO_TCP_STATE] &&
1259 nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]) >= TCP_CONNTRACK_MAX)
1262 spin_lock_bh(&ct->lock);
1263 if (tb[CTA_PROTOINFO_TCP_STATE])
1264 ct->proto.tcp.state = nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]);
1266 if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]) {
1267 struct nf_ct_tcp_flags *attr =
1268 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]);
1269 ct->proto.tcp.seen[0].flags &= ~attr->mask;
1270 ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask;
1273 if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]) {
1274 struct nf_ct_tcp_flags *attr =
1275 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]);
1276 ct->proto.tcp.seen[1].flags &= ~attr->mask;
1277 ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask;
1280 if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] &&
1281 tb[CTA_PROTOINFO_TCP_WSCALE_REPLY] &&
1282 ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_WINDOW_SCALE &&
1283 ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
1284 ct->proto.tcp.seen[0].td_scale =
1285 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL]);
1286 ct->proto.tcp.seen[1].td_scale =
1287 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_REPLY]);
1289 spin_unlock_bh(&ct->lock);
1294 static int tcp_nlattr_size(void)
1296 return nla_total_size(0) /* CTA_PROTOINFO_TCP */
1297 + nla_policy_len(tcp_nla_policy, CTA_PROTOINFO_TCP_MAX + 1);
1300 static int tcp_nlattr_tuple_size(void)
1302 return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1);
1306 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1308 #include <linux/netfilter/nfnetlink.h>
1309 #include <linux/netfilter/nfnetlink_cttimeout.h>
1311 static int tcp_timeout_nlattr_to_obj(struct nlattr *tb[],
1312 struct net *net, void *data)
1314 unsigned int *timeouts = data;
1315 struct nf_tcp_net *tn = tcp_pernet(net);
1318 /* set default TCP timeouts. */
1319 for (i=0; i<TCP_CONNTRACK_TIMEOUT_MAX; i++)
1320 timeouts[i] = tn->timeouts[i];
1322 if (tb[CTA_TIMEOUT_TCP_SYN_SENT]) {
1323 timeouts[TCP_CONNTRACK_SYN_SENT] =
1324 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT]))*HZ;
1326 if (tb[CTA_TIMEOUT_TCP_SYN_RECV]) {
1327 timeouts[TCP_CONNTRACK_SYN_RECV] =
1328 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_RECV]))*HZ;
1330 if (tb[CTA_TIMEOUT_TCP_ESTABLISHED]) {
1331 timeouts[TCP_CONNTRACK_ESTABLISHED] =
1332 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_ESTABLISHED]))*HZ;
1334 if (tb[CTA_TIMEOUT_TCP_FIN_WAIT]) {
1335 timeouts[TCP_CONNTRACK_FIN_WAIT] =
1336 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_FIN_WAIT]))*HZ;
1338 if (tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]) {
1339 timeouts[TCP_CONNTRACK_CLOSE_WAIT] =
1340 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]))*HZ;
1342 if (tb[CTA_TIMEOUT_TCP_LAST_ACK]) {
1343 timeouts[TCP_CONNTRACK_LAST_ACK] =
1344 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_LAST_ACK]))*HZ;
1346 if (tb[CTA_TIMEOUT_TCP_TIME_WAIT]) {
1347 timeouts[TCP_CONNTRACK_TIME_WAIT] =
1348 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_TIME_WAIT]))*HZ;
1350 if (tb[CTA_TIMEOUT_TCP_CLOSE]) {
1351 timeouts[TCP_CONNTRACK_CLOSE] =
1352 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE]))*HZ;
1354 if (tb[CTA_TIMEOUT_TCP_SYN_SENT2]) {
1355 timeouts[TCP_CONNTRACK_SYN_SENT2] =
1356 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT2]))*HZ;
1358 if (tb[CTA_TIMEOUT_TCP_RETRANS]) {
1359 timeouts[TCP_CONNTRACK_RETRANS] =
1360 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_RETRANS]))*HZ;
1362 if (tb[CTA_TIMEOUT_TCP_UNACK]) {
1363 timeouts[TCP_CONNTRACK_UNACK] =
1364 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_UNACK]))*HZ;
1370 tcp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
1372 const unsigned int *timeouts = data;
1374 if (nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT,
1375 htonl(timeouts[TCP_CONNTRACK_SYN_SENT] / HZ)) ||
1376 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_RECV,
1377 htonl(timeouts[TCP_CONNTRACK_SYN_RECV] / HZ)) ||
1378 nla_put_be32(skb, CTA_TIMEOUT_TCP_ESTABLISHED,
1379 htonl(timeouts[TCP_CONNTRACK_ESTABLISHED] / HZ)) ||
1380 nla_put_be32(skb, CTA_TIMEOUT_TCP_FIN_WAIT,
1381 htonl(timeouts[TCP_CONNTRACK_FIN_WAIT] / HZ)) ||
1382 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE_WAIT,
1383 htonl(timeouts[TCP_CONNTRACK_CLOSE_WAIT] / HZ)) ||
1384 nla_put_be32(skb, CTA_TIMEOUT_TCP_LAST_ACK,
1385 htonl(timeouts[TCP_CONNTRACK_LAST_ACK] / HZ)) ||
1386 nla_put_be32(skb, CTA_TIMEOUT_TCP_TIME_WAIT,
1387 htonl(timeouts[TCP_CONNTRACK_TIME_WAIT] / HZ)) ||
1388 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE,
1389 htonl(timeouts[TCP_CONNTRACK_CLOSE] / HZ)) ||
1390 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT2,
1391 htonl(timeouts[TCP_CONNTRACK_SYN_SENT2] / HZ)) ||
1392 nla_put_be32(skb, CTA_TIMEOUT_TCP_RETRANS,
1393 htonl(timeouts[TCP_CONNTRACK_RETRANS] / HZ)) ||
1394 nla_put_be32(skb, CTA_TIMEOUT_TCP_UNACK,
1395 htonl(timeouts[TCP_CONNTRACK_UNACK] / HZ)))
1396 goto nla_put_failure;
1403 static const struct nla_policy tcp_timeout_nla_policy[CTA_TIMEOUT_TCP_MAX+1] = {
1404 [CTA_TIMEOUT_TCP_SYN_SENT] = { .type = NLA_U32 },
1405 [CTA_TIMEOUT_TCP_SYN_RECV] = { .type = NLA_U32 },
1406 [CTA_TIMEOUT_TCP_ESTABLISHED] = { .type = NLA_U32 },
1407 [CTA_TIMEOUT_TCP_FIN_WAIT] = { .type = NLA_U32 },
1408 [CTA_TIMEOUT_TCP_CLOSE_WAIT] = { .type = NLA_U32 },
1409 [CTA_TIMEOUT_TCP_LAST_ACK] = { .type = NLA_U32 },
1410 [CTA_TIMEOUT_TCP_TIME_WAIT] = { .type = NLA_U32 },
1411 [CTA_TIMEOUT_TCP_CLOSE] = { .type = NLA_U32 },
1412 [CTA_TIMEOUT_TCP_SYN_SENT2] = { .type = NLA_U32 },
1413 [CTA_TIMEOUT_TCP_RETRANS] = { .type = NLA_U32 },
1414 [CTA_TIMEOUT_TCP_UNACK] = { .type = NLA_U32 },
1416 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1418 #ifdef CONFIG_SYSCTL
1419 static struct ctl_table tcp_sysctl_table[] = {
1421 .procname = "nf_conntrack_tcp_timeout_syn_sent",
1422 .maxlen = sizeof(unsigned int),
1424 .proc_handler = proc_dointvec_jiffies,
1427 .procname = "nf_conntrack_tcp_timeout_syn_recv",
1428 .maxlen = sizeof(unsigned int),
1430 .proc_handler = proc_dointvec_jiffies,
1433 .procname = "nf_conntrack_tcp_timeout_established",
1434 .maxlen = sizeof(unsigned int),
1436 .proc_handler = proc_dointvec_jiffies,
1439 .procname = "nf_conntrack_tcp_timeout_fin_wait",
1440 .maxlen = sizeof(unsigned int),
1442 .proc_handler = proc_dointvec_jiffies,
1445 .procname = "nf_conntrack_tcp_timeout_close_wait",
1446 .maxlen = sizeof(unsigned int),
1448 .proc_handler = proc_dointvec_jiffies,
1451 .procname = "nf_conntrack_tcp_timeout_last_ack",
1452 .maxlen = sizeof(unsigned int),
1454 .proc_handler = proc_dointvec_jiffies,
1457 .procname = "nf_conntrack_tcp_timeout_time_wait",
1458 .maxlen = sizeof(unsigned int),
1460 .proc_handler = proc_dointvec_jiffies,
1463 .procname = "nf_conntrack_tcp_timeout_close",
1464 .maxlen = sizeof(unsigned int),
1466 .proc_handler = proc_dointvec_jiffies,
1469 .procname = "nf_conntrack_tcp_timeout_max_retrans",
1470 .maxlen = sizeof(unsigned int),
1472 .proc_handler = proc_dointvec_jiffies,
1475 .procname = "nf_conntrack_tcp_timeout_unacknowledged",
1476 .maxlen = sizeof(unsigned int),
1478 .proc_handler = proc_dointvec_jiffies,
1481 .procname = "nf_conntrack_tcp_loose",
1482 .maxlen = sizeof(unsigned int),
1484 .proc_handler = proc_dointvec,
1487 .procname = "nf_conntrack_tcp_be_liberal",
1488 .maxlen = sizeof(unsigned int),
1490 .proc_handler = proc_dointvec,
1493 .procname = "nf_conntrack_tcp_max_retrans",
1494 .maxlen = sizeof(unsigned int),
1496 .proc_handler = proc_dointvec,
1500 #endif /* CONFIG_SYSCTL */
1502 static int tcp_kmemdup_sysctl_table(struct nf_proto_net *pn,
1503 struct nf_tcp_net *tn)
1505 #ifdef CONFIG_SYSCTL
1509 pn->ctl_table = kmemdup(tcp_sysctl_table,
1510 sizeof(tcp_sysctl_table),
1515 pn->ctl_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1516 pn->ctl_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1517 pn->ctl_table[2].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1518 pn->ctl_table[3].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1519 pn->ctl_table[4].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1520 pn->ctl_table[5].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1521 pn->ctl_table[6].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1522 pn->ctl_table[7].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1523 pn->ctl_table[8].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1524 pn->ctl_table[9].data = &tn->timeouts[TCP_CONNTRACK_UNACK];
1525 pn->ctl_table[10].data = &tn->tcp_loose;
1526 pn->ctl_table[11].data = &tn->tcp_be_liberal;
1527 pn->ctl_table[12].data = &tn->tcp_max_retrans;
1532 static int tcp_init_net(struct net *net, u_int16_t proto)
1534 struct nf_tcp_net *tn = tcp_pernet(net);
1535 struct nf_proto_net *pn = &tn->pn;
1540 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
1541 tn->timeouts[i] = tcp_timeouts[i];
1543 tn->tcp_loose = nf_ct_tcp_loose;
1544 tn->tcp_be_liberal = nf_ct_tcp_be_liberal;
1545 tn->tcp_max_retrans = nf_ct_tcp_max_retrans;
1548 return tcp_kmemdup_sysctl_table(pn, tn);
1551 static struct nf_proto_net *tcp_get_net_proto(struct net *net)
1553 return &net->ct.nf_ct_proto.tcp.pn;
1556 struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp4 __read_mostly =
1559 .l4proto = IPPROTO_TCP,
1560 .pkt_to_tuple = tcp_pkt_to_tuple,
1561 .invert_tuple = tcp_invert_tuple,
1562 #ifdef CONFIG_NF_CONNTRACK_PROCFS
1563 .print_conntrack = tcp_print_conntrack,
1565 .packet = tcp_packet,
1566 .get_timeouts = tcp_get_timeouts,
1569 .can_early_drop = tcp_can_early_drop,
1570 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1571 .to_nlattr = tcp_to_nlattr,
1572 .nlattr_size = tcp_nlattr_size,
1573 .from_nlattr = nlattr_to_tcp,
1574 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1575 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
1576 .nlattr_tuple_size = tcp_nlattr_tuple_size,
1577 .nla_policy = nf_ct_port_nla_policy,
1579 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1581 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1582 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1583 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1584 .obj_size = sizeof(unsigned int) *
1585 TCP_CONNTRACK_TIMEOUT_MAX,
1586 .nla_policy = tcp_timeout_nla_policy,
1588 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1589 .init_net = tcp_init_net,
1590 .get_net_proto = tcp_get_net_proto,
1592 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp4);
1594 struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6 __read_mostly =
1596 .l3proto = PF_INET6,
1597 .l4proto = IPPROTO_TCP,
1598 .pkt_to_tuple = tcp_pkt_to_tuple,
1599 .invert_tuple = tcp_invert_tuple,
1600 #ifdef CONFIG_NF_CONNTRACK_PROCFS
1601 .print_conntrack = tcp_print_conntrack,
1603 .packet = tcp_packet,
1604 .get_timeouts = tcp_get_timeouts,
1607 .can_early_drop = tcp_can_early_drop,
1608 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1609 .to_nlattr = tcp_to_nlattr,
1610 .nlattr_size = tcp_nlattr_size,
1611 .from_nlattr = nlattr_to_tcp,
1612 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1613 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
1614 .nlattr_tuple_size = tcp_nlattr_tuple_size,
1615 .nla_policy = nf_ct_port_nla_policy,
1617 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1619 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1620 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1621 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1622 .obj_size = sizeof(unsigned int) *
1623 TCP_CONNTRACK_TIMEOUT_MAX,
1624 .nla_policy = tcp_timeout_nla_policy,
1626 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1627 .init_net = tcp_init_net,
1628 .get_net_proto = tcp_get_net_proto,
1630 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp6);