d3d5a7fd73ce7c3fc79d4009cf4f0c1671ca2caf
[linux-2.6.git] / net / netfilter / nf_conntrack_proto_dccp.c
1 /*
2  * DCCP connection tracking protocol helper
3  *
4  * Copyright (c) 2005, 2006, 2008 Patrick McHardy <kaber@trash.net>
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License version 2 as
8  * published by the Free Software Foundation.
9  *
10  */
11 #include <linux/kernel.h>
12 #include <linux/module.h>
13 #include <linux/init.h>
14 #include <linux/sysctl.h>
15 #include <linux/spinlock.h>
16 #include <linux/skbuff.h>
17 #include <linux/dccp.h>
18
19 #include <net/net_namespace.h>
20 #include <net/netns/generic.h>
21
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_log.h>
26
27 static DEFINE_RWLOCK(dccp_lock);
28
29 /* Timeouts are based on values from RFC4340:
30  *
31  * - REQUEST:
32  *
33  *   8.1.2. Client Request
34  *
35  *   A client MAY give up on its DCCP-Requests after some time
36  *   (3 minutes, for example).
37  *
38  * - RESPOND:
39  *
40  *   8.1.3. Server Response
41  *
42  *   It MAY also leave the RESPOND state for CLOSED after a timeout of
43  *   not less than 4MSL (8 minutes);
44  *
45  * - PARTOPEN:
46  *
47  *   8.1.5. Handshake Completion
48  *
49  *   If the client remains in PARTOPEN for more than 4MSL (8 minutes),
50  *   it SHOULD reset the connection with Reset Code 2, "Aborted".
51  *
52  * - OPEN:
53  *
54  *   The DCCP timestamp overflows after 11.9 hours. If the connection
55  *   stays idle this long the sequence number won't be recognized
56  *   as valid anymore.
57  *
58  * - CLOSEREQ/CLOSING:
59  *
60  *   8.3. Termination
61  *
62  *   The retransmission timer should initially be set to go off in two
63  *   round-trip times and should back off to not less than once every
64  *   64 seconds ...
65  *
66  * - TIMEWAIT:
67  *
68  *   4.3. States
69  *
70  *   A server or client socket remains in this state for 2MSL (4 minutes)
71  *   after the connection has been town down, ...
72  */
73
74 #define DCCP_MSL (2 * 60 * HZ)
75
76 static const char * const dccp_state_names[] = {
77         [CT_DCCP_NONE]          = "NONE",
78         [CT_DCCP_REQUEST]       = "REQUEST",
79         [CT_DCCP_RESPOND]       = "RESPOND",
80         [CT_DCCP_PARTOPEN]      = "PARTOPEN",
81         [CT_DCCP_OPEN]          = "OPEN",
82         [CT_DCCP_CLOSEREQ]      = "CLOSEREQ",
83         [CT_DCCP_CLOSING]       = "CLOSING",
84         [CT_DCCP_TIMEWAIT]      = "TIMEWAIT",
85         [CT_DCCP_IGNORE]        = "IGNORE",
86         [CT_DCCP_INVALID]       = "INVALID",
87 };
88
89 #define sNO     CT_DCCP_NONE
90 #define sRQ     CT_DCCP_REQUEST
91 #define sRS     CT_DCCP_RESPOND
92 #define sPO     CT_DCCP_PARTOPEN
93 #define sOP     CT_DCCP_OPEN
94 #define sCR     CT_DCCP_CLOSEREQ
95 #define sCG     CT_DCCP_CLOSING
96 #define sTW     CT_DCCP_TIMEWAIT
97 #define sIG     CT_DCCP_IGNORE
98 #define sIV     CT_DCCP_INVALID
99
100 /*
101  * DCCP state transistion table
102  *
103  * The assumption is the same as for TCP tracking:
104  *
105  * We are the man in the middle. All the packets go through us but might
106  * get lost in transit to the destination. It is assumed that the destination
107  * can't receive segments we haven't seen.
108  *
109  * The following states exist:
110  *
111  * NONE:        Initial state, expecting Request
112  * REQUEST:     Request seen, waiting for Response from server
113  * RESPOND:     Response from server seen, waiting for Ack from client
114  * PARTOPEN:    Ack after Response seen, waiting for packet other than Response,
115  *              Reset or Sync from server
116  * OPEN:        Packet other than Response, Reset or Sync seen
117  * CLOSEREQ:    CloseReq from server seen, expecting Close from client
118  * CLOSING:     Close seen, expecting Reset
119  * TIMEWAIT:    Reset seen
120  * IGNORE:      Not determinable whether packet is valid
121  *
122  * Some states exist only on one side of the connection: REQUEST, RESPOND,
123  * PARTOPEN, CLOSEREQ. For the other side these states are equivalent to
124  * the one it was in before.
125  *
126  * Packets are marked as ignored (sIG) if we don't know if they're valid
127  * (for example a reincarnation of a connection we didn't notice is dead
128  * already) and the server may send back a connection closing Reset or a
129  * Response. They're also used for Sync/SyncAck packets, which we don't
130  * care about.
131  */
132 static const u_int8_t
133 dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] = {
134         [CT_DCCP_ROLE_CLIENT] = {
135                 [DCCP_PKT_REQUEST] = {
136                 /*
137                  * sNO -> sRQ           Regular Request
138                  * sRQ -> sRQ           Retransmitted Request or reincarnation
139                  * sRS -> sRS           Retransmitted Request (apparently Response
140                  *                      got lost after we saw it) or reincarnation
141                  * sPO -> sIG           Ignore, conntrack might be out of sync
142                  * sOP -> sIG           Ignore, conntrack might be out of sync
143                  * sCR -> sIG           Ignore, conntrack might be out of sync
144                  * sCG -> sIG           Ignore, conntrack might be out of sync
145                  * sTW -> sRQ           Reincarnation
146                  *
147                  *      sNO, sRQ, sRS, sPO. sOP, sCR, sCG, sTW, */
148                         sRQ, sRQ, sRS, sIG, sIG, sIG, sIG, sRQ,
149                 },
150                 [DCCP_PKT_RESPONSE] = {
151                 /*
152                  * sNO -> sIV           Invalid
153                  * sRQ -> sIG           Ignore, might be response to ignored Request
154                  * sRS -> sIG           Ignore, might be response to ignored Request
155                  * sPO -> sIG           Ignore, might be response to ignored Request
156                  * sOP -> sIG           Ignore, might be response to ignored Request
157                  * sCR -> sIG           Ignore, might be response to ignored Request
158                  * sCG -> sIG           Ignore, might be response to ignored Request
159                  * sTW -> sIV           Invalid, reincarnation in reverse direction
160                  *                      goes through sRQ
161                  *
162                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
163                         sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIV,
164                 },
165                 [DCCP_PKT_ACK] = {
166                 /*
167                  * sNO -> sIV           No connection
168                  * sRQ -> sIV           No connection
169                  * sRS -> sPO           Ack for Response, move to PARTOPEN (8.1.5.)
170                  * sPO -> sPO           Retransmitted Ack for Response, remain in PARTOPEN
171                  * sOP -> sOP           Regular ACK, remain in OPEN
172                  * sCR -> sCR           Ack in CLOSEREQ MAY be processed (8.3.)
173                  * sCG -> sCG           Ack in CLOSING MAY be processed (8.3.)
174                  * sTW -> sIV
175                  *
176                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
177                         sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
178                 },
179                 [DCCP_PKT_DATA] = {
180                 /*
181                  * sNO -> sIV           No connection
182                  * sRQ -> sIV           No connection
183                  * sRS -> sIV           No connection
184                  * sPO -> sIV           MUST use DataAck in PARTOPEN state (8.1.5.)
185                  * sOP -> sOP           Regular Data packet
186                  * sCR -> sCR           Data in CLOSEREQ MAY be processed (8.3.)
187                  * sCG -> sCG           Data in CLOSING MAY be processed (8.3.)
188                  * sTW -> sIV
189                  *
190                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
191                         sIV, sIV, sIV, sIV, sOP, sCR, sCG, sIV,
192                 },
193                 [DCCP_PKT_DATAACK] = {
194                 /*
195                  * sNO -> sIV           No connection
196                  * sRQ -> sIV           No connection
197                  * sRS -> sPO           Ack for Response, move to PARTOPEN (8.1.5.)
198                  * sPO -> sPO           Remain in PARTOPEN state
199                  * sOP -> sOP           Regular DataAck packet in OPEN state
200                  * sCR -> sCR           DataAck in CLOSEREQ MAY be processed (8.3.)
201                  * sCG -> sCG           DataAck in CLOSING MAY be processed (8.3.)
202                  * sTW -> sIV
203                  *
204                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
205                         sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
206                 },
207                 [DCCP_PKT_CLOSEREQ] = {
208                 /*
209                  * CLOSEREQ may only be sent by the server.
210                  *
211                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
212                         sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV
213                 },
214                 [DCCP_PKT_CLOSE] = {
215                 /*
216                  * sNO -> sIV           No connection
217                  * sRQ -> sIV           No connection
218                  * sRS -> sIV           No connection
219                  * sPO -> sCG           Client-initiated close
220                  * sOP -> sCG           Client-initiated close
221                  * sCR -> sCG           Close in response to CloseReq (8.3.)
222                  * sCG -> sCG           Retransmit
223                  * sTW -> sIV           Late retransmit, already in TIME_WAIT
224                  *
225                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
226                         sIV, sIV, sIV, sCG, sCG, sCG, sIV, sIV
227                 },
228                 [DCCP_PKT_RESET] = {
229                 /*
230                  * sNO -> sIV           No connection
231                  * sRQ -> sTW           Sync received or timeout, SHOULD send Reset (8.1.1.)
232                  * sRS -> sTW           Response received without Request
233                  * sPO -> sTW           Timeout, SHOULD send Reset (8.1.5.)
234                  * sOP -> sTW           Connection reset
235                  * sCR -> sTW           Connection reset
236                  * sCG -> sTW           Connection reset
237                  * sTW -> sIG           Ignore (don't refresh timer)
238                  *
239                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
240                         sIV, sTW, sTW, sTW, sTW, sTW, sTW, sIG
241                 },
242                 [DCCP_PKT_SYNC] = {
243                 /*
244                  * We currently ignore Sync packets
245                  *
246                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
247                         sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
248                 },
249                 [DCCP_PKT_SYNCACK] = {
250                 /*
251                  * We currently ignore SyncAck packets
252                  *
253                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
254                         sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
255                 },
256         },
257         [CT_DCCP_ROLE_SERVER] = {
258                 [DCCP_PKT_REQUEST] = {
259                 /*
260                  * sNO -> sIV           Invalid
261                  * sRQ -> sIG           Ignore, conntrack might be out of sync
262                  * sRS -> sIG           Ignore, conntrack might be out of sync
263                  * sPO -> sIG           Ignore, conntrack might be out of sync
264                  * sOP -> sIG           Ignore, conntrack might be out of sync
265                  * sCR -> sIG           Ignore, conntrack might be out of sync
266                  * sCG -> sIG           Ignore, conntrack might be out of sync
267                  * sTW -> sRQ           Reincarnation, must reverse roles
268                  *
269                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
270                         sIV, sIG, sIG, sIG, sIG, sIG, sIG, sRQ
271                 },
272                 [DCCP_PKT_RESPONSE] = {
273                 /*
274                  * sNO -> sIV           Response without Request
275                  * sRQ -> sRS           Response to clients Request
276                  * sRS -> sRS           Retransmitted Response (8.1.3. SHOULD NOT)
277                  * sPO -> sIG           Response to an ignored Request or late retransmit
278                  * sOP -> sIG           Ignore, might be response to ignored Request
279                  * sCR -> sIG           Ignore, might be response to ignored Request
280                  * sCG -> sIG           Ignore, might be response to ignored Request
281                  * sTW -> sIV           Invalid, Request from client in sTW moves to sRQ
282                  *
283                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
284                         sIV, sRS, sRS, sIG, sIG, sIG, sIG, sIV
285                 },
286                 [DCCP_PKT_ACK] = {
287                 /*
288                  * sNO -> sIV           No connection
289                  * sRQ -> sIV           No connection
290                  * sRS -> sIV           No connection
291                  * sPO -> sOP           Enter OPEN state (8.1.5.)
292                  * sOP -> sOP           Regular Ack in OPEN state
293                  * sCR -> sIV           Waiting for Close from client
294                  * sCG -> sCG           Ack in CLOSING MAY be processed (8.3.)
295                  * sTW -> sIV
296                  *
297                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
298                         sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
299                 },
300                 [DCCP_PKT_DATA] = {
301                 /*
302                  * sNO -> sIV           No connection
303                  * sRQ -> sIV           No connection
304                  * sRS -> sIV           No connection
305                  * sPO -> sOP           Enter OPEN state (8.1.5.)
306                  * sOP -> sOP           Regular Data packet in OPEN state
307                  * sCR -> sIV           Waiting for Close from client
308                  * sCG -> sCG           Data in CLOSING MAY be processed (8.3.)
309                  * sTW -> sIV
310                  *
311                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
312                         sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
313                 },
314                 [DCCP_PKT_DATAACK] = {
315                 /*
316                  * sNO -> sIV           No connection
317                  * sRQ -> sIV           No connection
318                  * sRS -> sIV           No connection
319                  * sPO -> sOP           Enter OPEN state (8.1.5.)
320                  * sOP -> sOP           Regular DataAck in OPEN state
321                  * sCR -> sIV           Waiting for Close from client
322                  * sCG -> sCG           Data in CLOSING MAY be processed (8.3.)
323                  * sTW -> sIV
324                  *
325                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
326                         sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
327                 },
328                 [DCCP_PKT_CLOSEREQ] = {
329                 /*
330                  * sNO -> sIV           No connection
331                  * sRQ -> sIV           No connection
332                  * sRS -> sIV           No connection
333                  * sPO -> sOP -> sCR    Move directly to CLOSEREQ (8.1.5.)
334                  * sOP -> sCR           CloseReq in OPEN state
335                  * sCR -> sCR           Retransmit
336                  * sCG -> sCR           Simultaneous close, client sends another Close
337                  * sTW -> sIV           Already closed
338                  *
339                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
340                         sIV, sIV, sIV, sCR, sCR, sCR, sCR, sIV
341                 },
342                 [DCCP_PKT_CLOSE] = {
343                 /*
344                  * sNO -> sIV           No connection
345                  * sRQ -> sIV           No connection
346                  * sRS -> sIV           No connection
347                  * sPO -> sOP -> sCG    Move direcly to CLOSING
348                  * sOP -> sCG           Move to CLOSING
349                  * sCR -> sIV           Close after CloseReq is invalid
350                  * sCG -> sCG           Retransmit
351                  * sTW -> sIV           Already closed
352                  *
353                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
354                         sIV, sIV, sIV, sCG, sCG, sIV, sCG, sIV
355                 },
356                 [DCCP_PKT_RESET] = {
357                 /*
358                  * sNO -> sIV           No connection
359                  * sRQ -> sTW           Reset in response to Request
360                  * sRS -> sTW           Timeout, SHOULD send Reset (8.1.3.)
361                  * sPO -> sTW           Timeout, SHOULD send Reset (8.1.3.)
362                  * sOP -> sTW
363                  * sCR -> sTW
364                  * sCG -> sTW
365                  * sTW -> sIG           Ignore (don't refresh timer)
366                  *
367                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW, sTW */
368                         sIV, sTW, sTW, sTW, sTW, sTW, sTW, sTW, sIG
369                 },
370                 [DCCP_PKT_SYNC] = {
371                 /*
372                  * We currently ignore Sync packets
373                  *
374                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
375                         sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
376                 },
377                 [DCCP_PKT_SYNCACK] = {
378                 /*
379                  * We currently ignore SyncAck packets
380                  *
381                  *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
382                         sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
383                 },
384         },
385 };
386
387 /* this module per-net specifics */
388 static int dccp_net_id;
389 struct dccp_net {
390         int dccp_loose;
391         unsigned int dccp_timeout[CT_DCCP_MAX + 1];
392 #ifdef CONFIG_SYSCTL
393         struct ctl_table_header *sysctl_header;
394         struct ctl_table *sysctl_table;
395 #endif
396 };
397
398 static inline struct dccp_net *dccp_pernet(struct net *net)
399 {
400         return net_generic(net, dccp_net_id);
401 }
402
403 static bool dccp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
404                               struct nf_conntrack_tuple *tuple)
405 {
406         struct dccp_hdr _hdr, *dh;
407
408         dh = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
409         if (dh == NULL)
410                 return false;
411
412         tuple->src.u.dccp.port = dh->dccph_sport;
413         tuple->dst.u.dccp.port = dh->dccph_dport;
414         return true;
415 }
416
417 static bool dccp_invert_tuple(struct nf_conntrack_tuple *inv,
418                               const struct nf_conntrack_tuple *tuple)
419 {
420         inv->src.u.dccp.port = tuple->dst.u.dccp.port;
421         inv->dst.u.dccp.port = tuple->src.u.dccp.port;
422         return true;
423 }
424
425 static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
426                      unsigned int dataoff)
427 {
428         struct net *net = nf_ct_net(ct);
429         struct dccp_net *dn;
430         struct dccp_hdr _dh, *dh;
431         const char *msg;
432         u_int8_t state;
433
434         dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
435         BUG_ON(dh == NULL);
436
437         state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
438         switch (state) {
439         default:
440                 dn = dccp_pernet(net);
441                 if (dn->dccp_loose == 0) {
442                         msg = "nf_ct_dccp: not picking up existing connection ";
443                         goto out_invalid;
444                 }
445         case CT_DCCP_REQUEST:
446                 break;
447         case CT_DCCP_INVALID:
448                 msg = "nf_ct_dccp: invalid state transition ";
449                 goto out_invalid;
450         }
451
452         ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
453         ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
454         ct->proto.dccp.state = CT_DCCP_NONE;
455         return true;
456
457 out_invalid:
458         if (LOG_INVALID(net, IPPROTO_DCCP))
459                 nf_log_packet(nf_ct_l3num(ct), 0, skb, NULL, NULL, NULL, msg);
460         return false;
461 }
462
463 static u64 dccp_ack_seq(const struct dccp_hdr *dh)
464 {
465         const struct dccp_hdr_ack_bits *dhack;
466
467         dhack = (void *)dh + __dccp_basic_hdr_len(dh);
468         return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) +
469                      ntohl(dhack->dccph_ack_nr_low);
470 }
471
472 static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
473                        unsigned int dataoff, enum ip_conntrack_info ctinfo,
474                        u_int8_t pf, unsigned int hooknum)
475 {
476         struct net *net = nf_ct_net(ct);
477         struct dccp_net *dn;
478         enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
479         struct dccp_hdr _dh, *dh;
480         u_int8_t type, old_state, new_state;
481         enum ct_dccp_roles role;
482
483         dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
484         BUG_ON(dh == NULL);
485         type = dh->dccph_type;
486
487         if (type == DCCP_PKT_RESET &&
488             !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
489                 /* Tear down connection immediately if only reply is a RESET */
490                 nf_ct_kill_acct(ct, ctinfo, skb);
491                 return NF_ACCEPT;
492         }
493
494         write_lock_bh(&dccp_lock);
495
496         role = ct->proto.dccp.role[dir];
497         old_state = ct->proto.dccp.state;
498         new_state = dccp_state_table[role][type][old_state];
499
500         switch (new_state) {
501         case CT_DCCP_REQUEST:
502                 if (old_state == CT_DCCP_TIMEWAIT &&
503                     role == CT_DCCP_ROLE_SERVER) {
504                         /* Reincarnation in the reverse direction: reopen and
505                          * reverse client/server roles. */
506                         ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
507                         ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
508                 }
509                 break;
510         case CT_DCCP_RESPOND:
511                 if (old_state == CT_DCCP_REQUEST)
512                         ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
513                 break;
514         case CT_DCCP_PARTOPEN:
515                 if (old_state == CT_DCCP_RESPOND &&
516                     type == DCCP_PKT_ACK &&
517                     dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
518                         set_bit(IPS_ASSURED_BIT, &ct->status);
519                 break;
520         case CT_DCCP_IGNORE:
521                 /*
522                  * Connection tracking might be out of sync, so we ignore
523                  * packets that might establish a new connection and resync
524                  * if the server responds with a valid Response.
525                  */
526                 if (ct->proto.dccp.last_dir == !dir &&
527                     ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
528                     type == DCCP_PKT_RESPONSE) {
529                         ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
530                         ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
531                         ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
532                         new_state = CT_DCCP_RESPOND;
533                         break;
534                 }
535                 ct->proto.dccp.last_dir = dir;
536                 ct->proto.dccp.last_pkt = type;
537
538                 write_unlock_bh(&dccp_lock);
539                 if (LOG_INVALID(net, IPPROTO_DCCP))
540                         nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
541                                       "nf_ct_dccp: invalid packet ignored ");
542                 return NF_ACCEPT;
543         case CT_DCCP_INVALID:
544                 write_unlock_bh(&dccp_lock);
545                 if (LOG_INVALID(net, IPPROTO_DCCP))
546                         nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
547                                       "nf_ct_dccp: invalid state transition ");
548                 return -NF_ACCEPT;
549         }
550
551         ct->proto.dccp.last_dir = dir;
552         ct->proto.dccp.last_pkt = type;
553         ct->proto.dccp.state = new_state;
554         write_unlock_bh(&dccp_lock);
555
556         dn = dccp_pernet(net);
557         nf_ct_refresh_acct(ct, ctinfo, skb, dn->dccp_timeout[new_state]);
558
559         return NF_ACCEPT;
560 }
561
562 static int dccp_error(struct net *net, struct sk_buff *skb,
563                       unsigned int dataoff, enum ip_conntrack_info *ctinfo,
564                       u_int8_t pf, unsigned int hooknum)
565 {
566         struct dccp_hdr _dh, *dh;
567         unsigned int dccp_len = skb->len - dataoff;
568         unsigned int cscov;
569         const char *msg;
570
571         dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
572         if (dh == NULL) {
573                 msg = "nf_ct_dccp: short packet ";
574                 goto out_invalid;
575         }
576
577         if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
578             dh->dccph_doff * 4 > dccp_len) {
579                 msg = "nf_ct_dccp: truncated/malformed packet ";
580                 goto out_invalid;
581         }
582
583         cscov = dccp_len;
584         if (dh->dccph_cscov) {
585                 cscov = (dh->dccph_cscov - 1) * 4;
586                 if (cscov > dccp_len) {
587                         msg = "nf_ct_dccp: bad checksum coverage ";
588                         goto out_invalid;
589                 }
590         }
591
592         if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
593             nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP,
594                                 pf)) {
595                 msg = "nf_ct_dccp: bad checksum ";
596                 goto out_invalid;
597         }
598
599         if (dh->dccph_type >= DCCP_PKT_INVALID) {
600                 msg = "nf_ct_dccp: reserved packet type ";
601                 goto out_invalid;
602         }
603
604         return NF_ACCEPT;
605
606 out_invalid:
607         if (LOG_INVALID(net, IPPROTO_DCCP))
608                 nf_log_packet(pf, 0, skb, NULL, NULL, NULL, msg);
609         return -NF_ACCEPT;
610 }
611
612 static int dccp_print_tuple(struct seq_file *s,
613                             const struct nf_conntrack_tuple *tuple)
614 {
615         return seq_printf(s, "sport=%hu dport=%hu ",
616                           ntohs(tuple->src.u.dccp.port),
617                           ntohs(tuple->dst.u.dccp.port));
618 }
619
620 static int dccp_print_conntrack(struct seq_file *s, const struct nf_conn *ct)
621 {
622         return seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
623 }
624
625 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
626 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
627                           const struct nf_conn *ct)
628 {
629         struct nlattr *nest_parms;
630
631         read_lock_bh(&dccp_lock);
632         nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED);
633         if (!nest_parms)
634                 goto nla_put_failure;
635         NLA_PUT_U8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state);
636         nla_nest_end(skb, nest_parms);
637         read_unlock_bh(&dccp_lock);
638         return 0;
639
640 nla_put_failure:
641         read_unlock_bh(&dccp_lock);
642         return -1;
643 }
644
645 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
646         [CTA_PROTOINFO_DCCP_STATE]      = { .type = NLA_U8 },
647 };
648
649 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
650 {
651         struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
652         struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
653         int err;
654
655         if (!attr)
656                 return 0;
657
658         err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr,
659                                dccp_nla_policy);
660         if (err < 0)
661                 return err;
662
663         if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
664             nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE)
665                 return -EINVAL;
666
667         write_lock_bh(&dccp_lock);
668         ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
669         write_unlock_bh(&dccp_lock);
670         return 0;
671 }
672 #endif
673
674 #ifdef CONFIG_SYSCTL
675 /* template, data assigned later */
676 static struct ctl_table dccp_sysctl_table[] = {
677         {
678                 .ctl_name       = CTL_UNNUMBERED,
679                 .procname       = "nf_conntrack_dccp_timeout_request",
680                 .maxlen         = sizeof(unsigned int),
681                 .mode           = 0644,
682                 .proc_handler   = proc_dointvec_jiffies,
683         },
684         {
685                 .ctl_name       = CTL_UNNUMBERED,
686                 .procname       = "nf_conntrack_dccp_timeout_respond",
687                 .maxlen         = sizeof(unsigned int),
688                 .mode           = 0644,
689                 .proc_handler   = proc_dointvec_jiffies,
690         },
691         {
692                 .ctl_name       = CTL_UNNUMBERED,
693                 .procname       = "nf_conntrack_dccp_timeout_partopen",
694                 .maxlen         = sizeof(unsigned int),
695                 .mode           = 0644,
696                 .proc_handler   = proc_dointvec_jiffies,
697         },
698         {
699                 .ctl_name       = CTL_UNNUMBERED,
700                 .procname       = "nf_conntrack_dccp_timeout_open",
701                 .maxlen         = sizeof(unsigned int),
702                 .mode           = 0644,
703                 .proc_handler   = proc_dointvec_jiffies,
704         },
705         {
706                 .ctl_name       = CTL_UNNUMBERED,
707                 .procname       = "nf_conntrack_dccp_timeout_closereq",
708                 .maxlen         = sizeof(unsigned int),
709                 .mode           = 0644,
710                 .proc_handler   = proc_dointvec_jiffies,
711         },
712         {
713                 .ctl_name       = CTL_UNNUMBERED,
714                 .procname       = "nf_conntrack_dccp_timeout_closing",
715                 .maxlen         = sizeof(unsigned int),
716                 .mode           = 0644,
717                 .proc_handler   = proc_dointvec_jiffies,
718         },
719         {
720                 .ctl_name       = CTL_UNNUMBERED,
721                 .procname       = "nf_conntrack_dccp_timeout_timewait",
722                 .maxlen         = sizeof(unsigned int),
723                 .mode           = 0644,
724                 .proc_handler   = proc_dointvec_jiffies,
725         },
726         {
727                 .ctl_name       = CTL_UNNUMBERED,
728                 .procname       = "nf_conntrack_dccp_loose",
729                 .maxlen         = sizeof(int),
730                 .mode           = 0644,
731                 .proc_handler   = proc_dointvec,
732         },
733         {
734                 .ctl_name       = 0,
735         }
736 };
737 #endif /* CONFIG_SYSCTL */
738
739 static struct nf_conntrack_l4proto dccp_proto4 __read_mostly = {
740         .l3proto                = AF_INET,
741         .l4proto                = IPPROTO_DCCP,
742         .name                   = "dccp",
743         .pkt_to_tuple           = dccp_pkt_to_tuple,
744         .invert_tuple           = dccp_invert_tuple,
745         .new                    = dccp_new,
746         .packet                 = dccp_packet,
747         .error                  = dccp_error,
748         .print_tuple            = dccp_print_tuple,
749         .print_conntrack        = dccp_print_conntrack,
750 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
751         .to_nlattr              = dccp_to_nlattr,
752         .from_nlattr            = nlattr_to_dccp,
753         .tuple_to_nlattr        = nf_ct_port_tuple_to_nlattr,
754         .nlattr_to_tuple        = nf_ct_port_nlattr_to_tuple,
755         .nla_policy             = nf_ct_port_nla_policy,
756 #endif
757 };
758
759 static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = {
760         .l3proto                = AF_INET6,
761         .l4proto                = IPPROTO_DCCP,
762         .name                   = "dccp",
763         .pkt_to_tuple           = dccp_pkt_to_tuple,
764         .invert_tuple           = dccp_invert_tuple,
765         .new                    = dccp_new,
766         .packet                 = dccp_packet,
767         .error                  = dccp_error,
768         .print_tuple            = dccp_print_tuple,
769         .print_conntrack        = dccp_print_conntrack,
770 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
771         .to_nlattr              = dccp_to_nlattr,
772         .from_nlattr            = nlattr_to_dccp,
773         .tuple_to_nlattr        = nf_ct_port_tuple_to_nlattr,
774         .nlattr_to_tuple        = nf_ct_port_nlattr_to_tuple,
775         .nla_policy             = nf_ct_port_nla_policy,
776 #endif
777 };
778
779 static __net_init int dccp_net_init(struct net *net)
780 {
781         struct dccp_net *dn;
782         int err;
783
784         dn = kmalloc(sizeof(*dn), GFP_KERNEL);
785         if (!dn)
786                 return -ENOMEM;
787
788         /* default values */
789         dn->dccp_loose = 1;
790         dn->dccp_timeout[CT_DCCP_REQUEST]       = 2 * DCCP_MSL;
791         dn->dccp_timeout[CT_DCCP_RESPOND]       = 4 * DCCP_MSL;
792         dn->dccp_timeout[CT_DCCP_PARTOPEN]      = 4 * DCCP_MSL;
793         dn->dccp_timeout[CT_DCCP_OPEN]          = 12 * 3600 * HZ;
794         dn->dccp_timeout[CT_DCCP_CLOSEREQ]      = 64 * HZ;
795         dn->dccp_timeout[CT_DCCP_CLOSING]       = 64 * HZ;
796         dn->dccp_timeout[CT_DCCP_TIMEWAIT]      = 2 * DCCP_MSL;
797
798         err = net_assign_generic(net, dccp_net_id, dn);
799         if (err)
800                 goto out;
801
802 #ifdef CONFIG_SYSCTL
803         err = -ENOMEM;
804         dn->sysctl_table = kmemdup(dccp_sysctl_table,
805                         sizeof(dccp_sysctl_table), GFP_KERNEL);
806         if (!dn->sysctl_table)
807                 goto out;
808
809         dn->sysctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST];
810         dn->sysctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND];
811         dn->sysctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN];
812         dn->sysctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN];
813         dn->sysctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ];
814         dn->sysctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING];
815         dn->sysctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT];
816         dn->sysctl_table[7].data = &dn->dccp_loose;
817
818         dn->sysctl_header = register_net_sysctl_table(net,
819                         nf_net_netfilter_sysctl_path, dn->sysctl_table);
820         if (!dn->sysctl_header) {
821                 kfree(dn->sysctl_table);
822                 goto out;
823         }
824 #endif
825
826         return 0;
827
828 out:
829         kfree(dn);
830         return err;
831 }
832
833 static __net_exit void dccp_net_exit(struct net *net)
834 {
835         struct dccp_net *dn = dccp_pernet(net);
836 #ifdef CONFIG_SYSCTL
837         unregister_net_sysctl_table(dn->sysctl_header);
838         kfree(dn->sysctl_table);
839 #endif
840         kfree(dn);
841
842         net_assign_generic(net, dccp_net_id, NULL);
843 }
844
845 static struct pernet_operations dccp_net_ops = {
846         .init = dccp_net_init,
847         .exit = dccp_net_exit,
848 };
849
850 static int __init nf_conntrack_proto_dccp_init(void)
851 {
852         int err;
853
854         err = register_pernet_gen_subsys(&dccp_net_id, &dccp_net_ops);
855         if (err < 0)
856                 goto err1;
857
858         err = nf_conntrack_l4proto_register(&dccp_proto4);
859         if (err < 0)
860                 goto err2;
861
862         err = nf_conntrack_l4proto_register(&dccp_proto6);
863         if (err < 0)
864                 goto err3;
865         return 0;
866
867 err3:
868         nf_conntrack_l4proto_unregister(&dccp_proto4);
869 err2:
870         unregister_pernet_gen_subsys(dccp_net_id, &dccp_net_ops);
871 err1:
872         return err;
873 }
874
875 static void __exit nf_conntrack_proto_dccp_fini(void)
876 {
877         unregister_pernet_gen_subsys(dccp_net_id, &dccp_net_ops);
878         nf_conntrack_l4proto_unregister(&dccp_proto6);
879         nf_conntrack_l4proto_unregister(&dccp_proto4);
880 }
881
882 module_init(nf_conntrack_proto_dccp_init);
883 module_exit(nf_conntrack_proto_dccp_fini);
884
885 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
886 MODULE_DESCRIPTION("DCCP connection tracking protocol helper");
887 MODULE_LICENSE("GPL");