2 * DCCP connection tracking protocol helper
4 * Copyright (c) 2005, 2006, 2008 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/kernel.h>
12 #include <linux/init.h>
13 #include <linux/sysctl.h>
14 #include <linux/spinlock.h>
15 #include <linux/skbuff.h>
16 #include <linux/dccp.h>
17 #include <linux/slab.h>
19 #include <net/net_namespace.h>
20 #include <net/netns/generic.h>
22 #include <linux/netfilter/nfnetlink_conntrack.h>
23 #include <net/netfilter/nf_conntrack.h>
24 #include <net/netfilter/nf_conntrack_l4proto.h>
25 #include <net/netfilter/nf_conntrack_ecache.h>
26 #include <net/netfilter/nf_log.h>
28 /* Timeouts are based on values from RFC4340:
32 * 8.1.2. Client Request
34 * A client MAY give up on its DCCP-Requests after some time
35 * (3 minutes, for example).
39 * 8.1.3. Server Response
41 * It MAY also leave the RESPOND state for CLOSED after a timeout of
42 * not less than 4MSL (8 minutes);
46 * 8.1.5. Handshake Completion
48 * If the client remains in PARTOPEN for more than 4MSL (8 minutes),
49 * it SHOULD reset the connection with Reset Code 2, "Aborted".
53 * The DCCP timestamp overflows after 11.9 hours. If the connection
54 * stays idle this long the sequence number won't be recognized
61 * The retransmission timer should initially be set to go off in two
62 * round-trip times and should back off to not less than once every
69 * A server or client socket remains in this state for 2MSL (4 minutes)
70 * after the connection has been town down, ...
73 #define DCCP_MSL (2 * 60 * HZ)
75 static const char * const dccp_state_names[] = {
76 [CT_DCCP_NONE] = "NONE",
77 [CT_DCCP_REQUEST] = "REQUEST",
78 [CT_DCCP_RESPOND] = "RESPOND",
79 [CT_DCCP_PARTOPEN] = "PARTOPEN",
80 [CT_DCCP_OPEN] = "OPEN",
81 [CT_DCCP_CLOSEREQ] = "CLOSEREQ",
82 [CT_DCCP_CLOSING] = "CLOSING",
83 [CT_DCCP_TIMEWAIT] = "TIMEWAIT",
84 [CT_DCCP_IGNORE] = "IGNORE",
85 [CT_DCCP_INVALID] = "INVALID",
88 #define sNO CT_DCCP_NONE
89 #define sRQ CT_DCCP_REQUEST
90 #define sRS CT_DCCP_RESPOND
91 #define sPO CT_DCCP_PARTOPEN
92 #define sOP CT_DCCP_OPEN
93 #define sCR CT_DCCP_CLOSEREQ
94 #define sCG CT_DCCP_CLOSING
95 #define sTW CT_DCCP_TIMEWAIT
96 #define sIG CT_DCCP_IGNORE
97 #define sIV CT_DCCP_INVALID
100 * DCCP state transition table
102 * The assumption is the same as for TCP tracking:
104 * We are the man in the middle. All the packets go through us but might
105 * get lost in transit to the destination. It is assumed that the destination
106 * can't receive segments we haven't seen.
108 * The following states exist:
110 * NONE: Initial state, expecting Request
111 * REQUEST: Request seen, waiting for Response from server
112 * RESPOND: Response from server seen, waiting for Ack from client
113 * PARTOPEN: Ack after Response seen, waiting for packet other than Response,
114 * Reset or Sync from server
115 * OPEN: Packet other than Response, Reset or Sync seen
116 * CLOSEREQ: CloseReq from server seen, expecting Close from client
117 * CLOSING: Close seen, expecting Reset
118 * TIMEWAIT: Reset seen
119 * IGNORE: Not determinable whether packet is valid
121 * Some states exist only on one side of the connection: REQUEST, RESPOND,
122 * PARTOPEN, CLOSEREQ. For the other side these states are equivalent to
123 * the one it was in before.
125 * Packets are marked as ignored (sIG) if we don't know if they're valid
126 * (for example a reincarnation of a connection we didn't notice is dead
127 * already) and the server may send back a connection closing Reset or a
128 * Response. They're also used for Sync/SyncAck packets, which we don't
131 static const u_int8_t
132 dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] = {
133 [CT_DCCP_ROLE_CLIENT] = {
134 [DCCP_PKT_REQUEST] = {
136 * sNO -> sRQ Regular Request
137 * sRQ -> sRQ Retransmitted Request or reincarnation
138 * sRS -> sRS Retransmitted Request (apparently Response
139 * got lost after we saw it) or reincarnation
140 * sPO -> sIG Ignore, conntrack might be out of sync
141 * sOP -> sIG Ignore, conntrack might be out of sync
142 * sCR -> sIG Ignore, conntrack might be out of sync
143 * sCG -> sIG Ignore, conntrack might be out of sync
144 * sTW -> sRQ Reincarnation
146 * sNO, sRQ, sRS, sPO. sOP, sCR, sCG, sTW, */
147 sRQ, sRQ, sRS, sIG, sIG, sIG, sIG, sRQ,
149 [DCCP_PKT_RESPONSE] = {
152 * sRQ -> sIG Ignore, might be response to ignored Request
153 * sRS -> sIG Ignore, might be response to ignored Request
154 * sPO -> sIG Ignore, might be response to ignored Request
155 * sOP -> sIG Ignore, might be response to ignored Request
156 * sCR -> sIG Ignore, might be response to ignored Request
157 * sCG -> sIG Ignore, might be response to ignored Request
158 * sTW -> sIV Invalid, reincarnation in reverse direction
161 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
162 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIV,
166 * sNO -> sIV No connection
167 * sRQ -> sIV No connection
168 * sRS -> sPO Ack for Response, move to PARTOPEN (8.1.5.)
169 * sPO -> sPO Retransmitted Ack for Response, remain in PARTOPEN
170 * sOP -> sOP Regular ACK, remain in OPEN
171 * sCR -> sCR Ack in CLOSEREQ MAY be processed (8.3.)
172 * sCG -> sCG Ack in CLOSING MAY be processed (8.3.)
175 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
176 sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
180 * sNO -> sIV No connection
181 * sRQ -> sIV No connection
182 * sRS -> sIV No connection
183 * sPO -> sIV MUST use DataAck in PARTOPEN state (8.1.5.)
184 * sOP -> sOP Regular Data packet
185 * sCR -> sCR Data in CLOSEREQ MAY be processed (8.3.)
186 * sCG -> sCG Data in CLOSING MAY be processed (8.3.)
189 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
190 sIV, sIV, sIV, sIV, sOP, sCR, sCG, sIV,
192 [DCCP_PKT_DATAACK] = {
194 * sNO -> sIV No connection
195 * sRQ -> sIV No connection
196 * sRS -> sPO Ack for Response, move to PARTOPEN (8.1.5.)
197 * sPO -> sPO Remain in PARTOPEN state
198 * sOP -> sOP Regular DataAck packet in OPEN state
199 * sCR -> sCR DataAck in CLOSEREQ MAY be processed (8.3.)
200 * sCG -> sCG DataAck in CLOSING MAY be processed (8.3.)
203 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
204 sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
206 [DCCP_PKT_CLOSEREQ] = {
208 * CLOSEREQ may only be sent by the server.
210 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
211 sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV
215 * sNO -> sIV No connection
216 * sRQ -> sIV No connection
217 * sRS -> sIV No connection
218 * sPO -> sCG Client-initiated close
219 * sOP -> sCG Client-initiated close
220 * sCR -> sCG Close in response to CloseReq (8.3.)
221 * sCG -> sCG Retransmit
222 * sTW -> sIV Late retransmit, already in TIME_WAIT
224 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
225 sIV, sIV, sIV, sCG, sCG, sCG, sIV, sIV
229 * sNO -> sIV No connection
230 * sRQ -> sTW Sync received or timeout, SHOULD send Reset (8.1.1.)
231 * sRS -> sTW Response received without Request
232 * sPO -> sTW Timeout, SHOULD send Reset (8.1.5.)
233 * sOP -> sTW Connection reset
234 * sCR -> sTW Connection reset
235 * sCG -> sTW Connection reset
236 * sTW -> sIG Ignore (don't refresh timer)
238 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
239 sIV, sTW, sTW, sTW, sTW, sTW, sTW, sIG
243 * We currently ignore Sync packets
245 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
246 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
248 [DCCP_PKT_SYNCACK] = {
250 * We currently ignore SyncAck packets
252 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
253 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
256 [CT_DCCP_ROLE_SERVER] = {
257 [DCCP_PKT_REQUEST] = {
260 * sRQ -> sIG Ignore, conntrack might be out of sync
261 * sRS -> sIG Ignore, conntrack might be out of sync
262 * sPO -> sIG Ignore, conntrack might be out of sync
263 * sOP -> sIG Ignore, conntrack might be out of sync
264 * sCR -> sIG Ignore, conntrack might be out of sync
265 * sCG -> sIG Ignore, conntrack might be out of sync
266 * sTW -> sRQ Reincarnation, must reverse roles
268 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
269 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sRQ
271 [DCCP_PKT_RESPONSE] = {
273 * sNO -> sIV Response without Request
274 * sRQ -> sRS Response to clients Request
275 * sRS -> sRS Retransmitted Response (8.1.3. SHOULD NOT)
276 * sPO -> sIG Response to an ignored Request or late retransmit
277 * sOP -> sIG Ignore, might be response to ignored Request
278 * sCR -> sIG Ignore, might be response to ignored Request
279 * sCG -> sIG Ignore, might be response to ignored Request
280 * sTW -> sIV Invalid, Request from client in sTW moves to sRQ
282 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
283 sIV, sRS, sRS, sIG, sIG, sIG, sIG, sIV
287 * sNO -> sIV No connection
288 * sRQ -> sIV No connection
289 * sRS -> sIV No connection
290 * sPO -> sOP Enter OPEN state (8.1.5.)
291 * sOP -> sOP Regular Ack in OPEN state
292 * sCR -> sIV Waiting for Close from client
293 * sCG -> sCG Ack in CLOSING MAY be processed (8.3.)
296 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
297 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
301 * sNO -> sIV No connection
302 * sRQ -> sIV No connection
303 * sRS -> sIV No connection
304 * sPO -> sOP Enter OPEN state (8.1.5.)
305 * sOP -> sOP Regular Data packet in OPEN state
306 * sCR -> sIV Waiting for Close from client
307 * sCG -> sCG Data in CLOSING MAY be processed (8.3.)
310 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
311 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
313 [DCCP_PKT_DATAACK] = {
315 * sNO -> sIV No connection
316 * sRQ -> sIV No connection
317 * sRS -> sIV No connection
318 * sPO -> sOP Enter OPEN state (8.1.5.)
319 * sOP -> sOP Regular DataAck in OPEN state
320 * sCR -> sIV Waiting for Close from client
321 * sCG -> sCG Data in CLOSING MAY be processed (8.3.)
324 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
325 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
327 [DCCP_PKT_CLOSEREQ] = {
329 * sNO -> sIV No connection
330 * sRQ -> sIV No connection
331 * sRS -> sIV No connection
332 * sPO -> sOP -> sCR Move directly to CLOSEREQ (8.1.5.)
333 * sOP -> sCR CloseReq in OPEN state
334 * sCR -> sCR Retransmit
335 * sCG -> sCR Simultaneous close, client sends another Close
336 * sTW -> sIV Already closed
338 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
339 sIV, sIV, sIV, sCR, sCR, sCR, sCR, sIV
343 * sNO -> sIV No connection
344 * sRQ -> sIV No connection
345 * sRS -> sIV No connection
346 * sPO -> sOP -> sCG Move direcly to CLOSING
347 * sOP -> sCG Move to CLOSING
348 * sCR -> sIV Close after CloseReq is invalid
349 * sCG -> sCG Retransmit
350 * sTW -> sIV Already closed
352 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
353 sIV, sIV, sIV, sCG, sCG, sIV, sCG, sIV
357 * sNO -> sIV No connection
358 * sRQ -> sTW Reset in response to Request
359 * sRS -> sTW Timeout, SHOULD send Reset (8.1.3.)
360 * sPO -> sTW Timeout, SHOULD send Reset (8.1.3.)
364 * sTW -> sIG Ignore (don't refresh timer)
366 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW, sTW */
367 sIV, sTW, sTW, sTW, sTW, sTW, sTW, sTW, sIG
371 * We currently ignore Sync packets
373 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
374 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
376 [DCCP_PKT_SYNCACK] = {
378 * We currently ignore SyncAck packets
380 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
381 sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
386 static inline struct nf_dccp_net *dccp_pernet(struct net *net)
388 return &net->ct.nf_ct_proto.dccp;
391 static bool dccp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
392 struct net *net, struct nf_conntrack_tuple *tuple)
394 struct dccp_hdr _hdr, *dh;
396 /* Actually only need first 4 bytes to get ports. */
397 dh = skb_header_pointer(skb, dataoff, 4, &_hdr);
401 tuple->src.u.dccp.port = dh->dccph_sport;
402 tuple->dst.u.dccp.port = dh->dccph_dport;
406 static bool dccp_invert_tuple(struct nf_conntrack_tuple *inv,
407 const struct nf_conntrack_tuple *tuple)
409 inv->src.u.dccp.port = tuple->dst.u.dccp.port;
410 inv->dst.u.dccp.port = tuple->src.u.dccp.port;
414 static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
415 unsigned int dataoff, unsigned int *timeouts)
417 struct net *net = nf_ct_net(ct);
418 struct nf_dccp_net *dn;
419 struct dccp_hdr _dh, *dh;
423 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh);
426 state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
429 dn = dccp_pernet(net);
430 if (dn->dccp_loose == 0) {
431 msg = "not picking up existing connection ";
434 case CT_DCCP_REQUEST:
436 case CT_DCCP_INVALID:
437 msg = "invalid state transition ";
441 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
442 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
443 ct->proto.dccp.state = CT_DCCP_NONE;
444 ct->proto.dccp.last_pkt = DCCP_PKT_REQUEST;
445 ct->proto.dccp.last_dir = IP_CT_DIR_ORIGINAL;
446 ct->proto.dccp.handshake_seq = 0;
450 nf_ct_l4proto_log_invalid(skb, ct, "%s", msg);
454 static u64 dccp_ack_seq(const struct dccp_hdr *dh)
456 const struct dccp_hdr_ack_bits *dhack;
458 dhack = (void *)dh + __dccp_basic_hdr_len(dh);
459 return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) +
460 ntohl(dhack->dccph_ack_nr_low);
463 static unsigned int *dccp_get_timeouts(struct net *net)
465 return dccp_pernet(net)->dccp_timeout;
468 static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
469 unsigned int dataoff, enum ip_conntrack_info ctinfo,
470 unsigned int *timeouts)
472 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
473 struct dccp_hdr _dh, *dh;
474 u_int8_t type, old_state, new_state;
475 enum ct_dccp_roles role;
477 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh);
479 type = dh->dccph_type;
481 if (type == DCCP_PKT_RESET &&
482 !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
483 /* Tear down connection immediately if only reply is a RESET */
484 nf_ct_kill_acct(ct, ctinfo, skb);
488 spin_lock_bh(&ct->lock);
490 role = ct->proto.dccp.role[dir];
491 old_state = ct->proto.dccp.state;
492 new_state = dccp_state_table[role][type][old_state];
495 case CT_DCCP_REQUEST:
496 if (old_state == CT_DCCP_TIMEWAIT &&
497 role == CT_DCCP_ROLE_SERVER) {
498 /* Reincarnation in the reverse direction: reopen and
499 * reverse client/server roles. */
500 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
501 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
504 case CT_DCCP_RESPOND:
505 if (old_state == CT_DCCP_REQUEST)
506 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
508 case CT_DCCP_PARTOPEN:
509 if (old_state == CT_DCCP_RESPOND &&
510 type == DCCP_PKT_ACK &&
511 dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
512 set_bit(IPS_ASSURED_BIT, &ct->status);
516 * Connection tracking might be out of sync, so we ignore
517 * packets that might establish a new connection and resync
518 * if the server responds with a valid Response.
520 if (ct->proto.dccp.last_dir == !dir &&
521 ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
522 type == DCCP_PKT_RESPONSE) {
523 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
524 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
525 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
526 new_state = CT_DCCP_RESPOND;
529 ct->proto.dccp.last_dir = dir;
530 ct->proto.dccp.last_pkt = type;
532 spin_unlock_bh(&ct->lock);
533 nf_ct_l4proto_log_invalid(skb, ct, "%s", "invalid packet");
535 case CT_DCCP_INVALID:
536 spin_unlock_bh(&ct->lock);
537 nf_ct_l4proto_log_invalid(skb, ct, "%s", "invalid state transition");
541 ct->proto.dccp.last_dir = dir;
542 ct->proto.dccp.last_pkt = type;
543 ct->proto.dccp.state = new_state;
544 spin_unlock_bh(&ct->lock);
546 if (new_state != old_state)
547 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
549 nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]);
554 static int dccp_error(struct net *net, struct nf_conn *tmpl,
555 struct sk_buff *skb, unsigned int dataoff,
556 u_int8_t pf, unsigned int hooknum)
558 struct dccp_hdr _dh, *dh;
559 unsigned int dccp_len = skb->len - dataoff;
563 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh);
565 msg = "nf_ct_dccp: short packet ";
569 if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
570 dh->dccph_doff * 4 > dccp_len) {
571 msg = "nf_ct_dccp: truncated/malformed packet ";
576 if (dh->dccph_cscov) {
577 cscov = (dh->dccph_cscov - 1) * 4;
578 if (cscov > dccp_len) {
579 msg = "nf_ct_dccp: bad checksum coverage ";
584 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
585 nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP,
587 msg = "nf_ct_dccp: bad checksum ";
591 if (dh->dccph_type >= DCCP_PKT_INVALID) {
592 msg = "nf_ct_dccp: reserved packet type ";
599 nf_l4proto_log_invalid(skb, net, pf, IPPROTO_DCCP, "%s", msg);
603 static bool dccp_can_early_drop(const struct nf_conn *ct)
605 switch (ct->proto.dccp.state) {
606 case CT_DCCP_CLOSEREQ:
607 case CT_DCCP_CLOSING:
608 case CT_DCCP_TIMEWAIT:
617 #ifdef CONFIG_NF_CONNTRACK_PROCFS
618 static void dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
620 seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
624 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
625 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
628 struct nlattr *nest_parms;
630 spin_lock_bh(&ct->lock);
631 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED);
633 goto nla_put_failure;
634 if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state) ||
635 nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE,
636 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) ||
637 nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ,
638 cpu_to_be64(ct->proto.dccp.handshake_seq),
639 CTA_PROTOINFO_DCCP_PAD))
640 goto nla_put_failure;
641 nla_nest_end(skb, nest_parms);
642 spin_unlock_bh(&ct->lock);
646 spin_unlock_bh(&ct->lock);
650 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
651 [CTA_PROTOINFO_DCCP_STATE] = { .type = NLA_U8 },
652 [CTA_PROTOINFO_DCCP_ROLE] = { .type = NLA_U8 },
653 [CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ] = { .type = NLA_U64 },
654 [CTA_PROTOINFO_DCCP_PAD] = { .type = NLA_UNSPEC },
657 #define DCCP_NLATTR_SIZE ( \
658 NLA_ALIGN(NLA_HDRLEN + 1) + \
659 NLA_ALIGN(NLA_HDRLEN + 1) + \
660 NLA_ALIGN(NLA_HDRLEN + sizeof(u64)) + \
661 NLA_ALIGN(NLA_HDRLEN + 0))
663 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
665 struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
666 struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
672 err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr,
673 dccp_nla_policy, NULL);
677 if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
678 !tb[CTA_PROTOINFO_DCCP_ROLE] ||
679 nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX ||
680 nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) {
684 spin_lock_bh(&ct->lock);
685 ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
686 if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) {
687 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
688 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
690 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_SERVER;
691 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_CLIENT;
693 if (tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]) {
694 ct->proto.dccp.handshake_seq =
695 be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]));
697 spin_unlock_bh(&ct->lock);
702 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
704 #include <linux/netfilter/nfnetlink.h>
705 #include <linux/netfilter/nfnetlink_cttimeout.h>
707 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[],
708 struct net *net, void *data)
710 struct nf_dccp_net *dn = dccp_pernet(net);
711 unsigned int *timeouts = data;
714 /* set default DCCP timeouts. */
715 for (i=0; i<CT_DCCP_MAX; i++)
716 timeouts[i] = dn->dccp_timeout[i];
718 /* there's a 1:1 mapping between attributes and protocol states. */
719 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
721 timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ;
728 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
730 const unsigned int *timeouts = data;
733 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
734 if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ)))
735 goto nla_put_failure;
743 static const struct nla_policy
744 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = {
745 [CTA_TIMEOUT_DCCP_REQUEST] = { .type = NLA_U32 },
746 [CTA_TIMEOUT_DCCP_RESPOND] = { .type = NLA_U32 },
747 [CTA_TIMEOUT_DCCP_PARTOPEN] = { .type = NLA_U32 },
748 [CTA_TIMEOUT_DCCP_OPEN] = { .type = NLA_U32 },
749 [CTA_TIMEOUT_DCCP_CLOSEREQ] = { .type = NLA_U32 },
750 [CTA_TIMEOUT_DCCP_CLOSING] = { .type = NLA_U32 },
751 [CTA_TIMEOUT_DCCP_TIMEWAIT] = { .type = NLA_U32 },
753 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
756 /* template, data assigned later */
757 static struct ctl_table dccp_sysctl_table[] = {
759 .procname = "nf_conntrack_dccp_timeout_request",
760 .maxlen = sizeof(unsigned int),
762 .proc_handler = proc_dointvec_jiffies,
765 .procname = "nf_conntrack_dccp_timeout_respond",
766 .maxlen = sizeof(unsigned int),
768 .proc_handler = proc_dointvec_jiffies,
771 .procname = "nf_conntrack_dccp_timeout_partopen",
772 .maxlen = sizeof(unsigned int),
774 .proc_handler = proc_dointvec_jiffies,
777 .procname = "nf_conntrack_dccp_timeout_open",
778 .maxlen = sizeof(unsigned int),
780 .proc_handler = proc_dointvec_jiffies,
783 .procname = "nf_conntrack_dccp_timeout_closereq",
784 .maxlen = sizeof(unsigned int),
786 .proc_handler = proc_dointvec_jiffies,
789 .procname = "nf_conntrack_dccp_timeout_closing",
790 .maxlen = sizeof(unsigned int),
792 .proc_handler = proc_dointvec_jiffies,
795 .procname = "nf_conntrack_dccp_timeout_timewait",
796 .maxlen = sizeof(unsigned int),
798 .proc_handler = proc_dointvec_jiffies,
801 .procname = "nf_conntrack_dccp_loose",
802 .maxlen = sizeof(int),
804 .proc_handler = proc_dointvec,
808 #endif /* CONFIG_SYSCTL */
810 static int dccp_kmemdup_sysctl_table(struct net *net, struct nf_proto_net *pn,
811 struct nf_dccp_net *dn)
817 pn->ctl_table = kmemdup(dccp_sysctl_table,
818 sizeof(dccp_sysctl_table),
823 pn->ctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST];
824 pn->ctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND];
825 pn->ctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN];
826 pn->ctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN];
827 pn->ctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ];
828 pn->ctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING];
829 pn->ctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT];
830 pn->ctl_table[7].data = &dn->dccp_loose;
832 /* Don't export sysctls to unprivileged users */
833 if (net->user_ns != &init_user_ns)
834 pn->ctl_table[0].procname = NULL;
839 static int dccp_init_net(struct net *net, u_int16_t proto)
841 struct nf_dccp_net *dn = dccp_pernet(net);
842 struct nf_proto_net *pn = &dn->pn;
847 dn->dccp_timeout[CT_DCCP_REQUEST] = 2 * DCCP_MSL;
848 dn->dccp_timeout[CT_DCCP_RESPOND] = 4 * DCCP_MSL;
849 dn->dccp_timeout[CT_DCCP_PARTOPEN] = 4 * DCCP_MSL;
850 dn->dccp_timeout[CT_DCCP_OPEN] = 12 * 3600 * HZ;
851 dn->dccp_timeout[CT_DCCP_CLOSEREQ] = 64 * HZ;
852 dn->dccp_timeout[CT_DCCP_CLOSING] = 64 * HZ;
853 dn->dccp_timeout[CT_DCCP_TIMEWAIT] = 2 * DCCP_MSL;
856 return dccp_kmemdup_sysctl_table(net, pn, dn);
859 static struct nf_proto_net *dccp_get_net_proto(struct net *net)
861 return &net->ct.nf_ct_proto.dccp.pn;
864 const struct nf_conntrack_l4proto nf_conntrack_l4proto_dccp4 = {
866 .l4proto = IPPROTO_DCCP,
867 .pkt_to_tuple = dccp_pkt_to_tuple,
868 .invert_tuple = dccp_invert_tuple,
870 .packet = dccp_packet,
871 .get_timeouts = dccp_get_timeouts,
873 .can_early_drop = dccp_can_early_drop,
874 #ifdef CONFIG_NF_CONNTRACK_PROCFS
875 .print_conntrack = dccp_print_conntrack,
877 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
878 .nlattr_size = DCCP_NLATTR_SIZE,
879 .to_nlattr = dccp_to_nlattr,
880 .from_nlattr = nlattr_to_dccp,
881 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
882 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
883 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
884 .nla_policy = nf_ct_port_nla_policy,
886 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
888 .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
889 .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
890 .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
891 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
892 .nla_policy = dccp_timeout_nla_policy,
894 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
895 .init_net = dccp_init_net,
896 .get_net_proto = dccp_get_net_proto,
898 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_dccp4);
900 const struct nf_conntrack_l4proto nf_conntrack_l4proto_dccp6 = {
902 .l4proto = IPPROTO_DCCP,
903 .pkt_to_tuple = dccp_pkt_to_tuple,
904 .invert_tuple = dccp_invert_tuple,
906 .packet = dccp_packet,
907 .get_timeouts = dccp_get_timeouts,
909 .can_early_drop = dccp_can_early_drop,
910 #ifdef CONFIG_NF_CONNTRACK_PROCFS
911 .print_conntrack = dccp_print_conntrack,
913 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
914 .nlattr_size = DCCP_NLATTR_SIZE,
915 .to_nlattr = dccp_to_nlattr,
916 .from_nlattr = nlattr_to_dccp,
917 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
918 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
919 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
920 .nla_policy = nf_ct_port_nla_policy,
922 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
924 .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
925 .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
926 .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
927 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
928 .nla_policy = dccp_timeout_nla_policy,
930 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
931 .init_net = dccp_init_net,
932 .get_net_proto = dccp_get_net_proto,
934 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_dccp6);