Linux Kernel  3.7.1
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
nf_conntrack_proto_dccp.c
Go to the documentation of this file.
1 /*
2  * DCCP connection tracking protocol helper
3  *
4  * Copyright (c) 2005, 2006, 2008 Patrick McHardy <[email protected]>
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 #include <linux/slab.h>
19 
20 #include <net/net_namespace.h>
21 #include <net/netns/generic.h>
22 
27 #include <net/netfilter/nf_log.h>
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 transition 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 __read_mostly;
389 struct dccp_net {
390  struct nf_proto_net pn;
392  unsigned int dccp_timeout[CT_DCCP_MAX + 1];
393 };
394 
395 static inline struct dccp_net *dccp_pernet(struct net *net)
396 {
397  return net_generic(net, dccp_net_id);
398 }
399 
400 static bool dccp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
401  struct nf_conntrack_tuple *tuple)
402 {
403  struct dccp_hdr _hdr, *dh;
404 
405  dh = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
406  if (dh == NULL)
407  return false;
408 
409  tuple->src.u.dccp.port = dh->dccph_sport;
410  tuple->dst.u.dccp.port = dh->dccph_dport;
411  return true;
412 }
413 
414 static bool dccp_invert_tuple(struct nf_conntrack_tuple *inv,
415  const struct nf_conntrack_tuple *tuple)
416 {
417  inv->src.u.dccp.port = tuple->dst.u.dccp.port;
418  inv->dst.u.dccp.port = tuple->src.u.dccp.port;
419  return true;
420 }
421 
422 static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
423  unsigned int dataoff, unsigned int *timeouts)
424 {
425  struct net *net = nf_ct_net(ct);
426  struct dccp_net *dn;
427  struct dccp_hdr _dh, *dh;
428  const char *msg;
429  u_int8_t state;
430 
431  dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
432  BUG_ON(dh == NULL);
433 
434  state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
435  switch (state) {
436  default:
437  dn = dccp_pernet(net);
438  if (dn->dccp_loose == 0) {
439  msg = "nf_ct_dccp: not picking up existing connection ";
440  goto out_invalid;
441  }
442  case CT_DCCP_REQUEST:
443  break;
444  case CT_DCCP_INVALID:
445  msg = "nf_ct_dccp: invalid state transition ";
446  goto out_invalid;
447  }
448 
451  ct->proto.dccp.state = CT_DCCP_NONE;
452  ct->proto.dccp.last_pkt = DCCP_PKT_REQUEST;
453  ct->proto.dccp.last_dir = IP_CT_DIR_ORIGINAL;
454  ct->proto.dccp.handshake_seq = 0;
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 unsigned int *dccp_get_timeouts(struct net *net)
473 {
474  return dccp_pernet(net)->dccp_timeout;
475 }
476 
477 static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
478  unsigned int dataoff, enum ip_conntrack_info ctinfo,
479  u_int8_t pf, unsigned int hooknum,
480  unsigned int *timeouts)
481 {
482  struct net *net = nf_ct_net(ct);
483  enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
484  struct dccp_hdr _dh, *dh;
485  u_int8_t type, old_state, new_state;
486  enum ct_dccp_roles role;
487 
488  dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
489  BUG_ON(dh == NULL);
490  type = dh->dccph_type;
491 
492  if (type == DCCP_PKT_RESET &&
494  /* Tear down connection immediately if only reply is a RESET */
495  nf_ct_kill_acct(ct, ctinfo, skb);
496  return NF_ACCEPT;
497  }
498 
499  spin_lock_bh(&ct->lock);
500 
501  role = ct->proto.dccp.role[dir];
502  old_state = ct->proto.dccp.state;
503  new_state = dccp_state_table[role][type][old_state];
504 
505  switch (new_state) {
506  case CT_DCCP_REQUEST:
507  if (old_state == CT_DCCP_TIMEWAIT &&
508  role == CT_DCCP_ROLE_SERVER) {
509  /* Reincarnation in the reverse direction: reopen and
510  * reverse client/server roles. */
511  ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
512  ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
513  }
514  break;
515  case CT_DCCP_RESPOND:
516  if (old_state == CT_DCCP_REQUEST)
517  ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
518  break;
519  case CT_DCCP_PARTOPEN:
520  if (old_state == CT_DCCP_RESPOND &&
521  type == DCCP_PKT_ACK &&
522  dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
524  break;
525  case CT_DCCP_IGNORE:
526  /*
527  * Connection tracking might be out of sync, so we ignore
528  * packets that might establish a new connection and resync
529  * if the server responds with a valid Response.
530  */
531  if (ct->proto.dccp.last_dir == !dir &&
532  ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
533  type == DCCP_PKT_RESPONSE) {
534  ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
535  ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
536  ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
537  new_state = CT_DCCP_RESPOND;
538  break;
539  }
540  ct->proto.dccp.last_dir = dir;
541  ct->proto.dccp.last_pkt = type;
542 
543  spin_unlock_bh(&ct->lock);
544  if (LOG_INVALID(net, IPPROTO_DCCP))
545  nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
546  "nf_ct_dccp: invalid packet ignored ");
547  return NF_ACCEPT;
548  case CT_DCCP_INVALID:
549  spin_unlock_bh(&ct->lock);
550  if (LOG_INVALID(net, IPPROTO_DCCP))
551  nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
552  "nf_ct_dccp: invalid state transition ");
553  return -NF_ACCEPT;
554  }
555 
556  ct->proto.dccp.last_dir = dir;
557  ct->proto.dccp.last_pkt = type;
558  ct->proto.dccp.state = new_state;
559  spin_unlock_bh(&ct->lock);
560 
561  if (new_state != old_state)
562  nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
563 
564  nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]);
565 
566  return NF_ACCEPT;
567 }
568 
569 static int dccp_error(struct net *net, struct nf_conn *tmpl,
570  struct sk_buff *skb, unsigned int dataoff,
571  enum ip_conntrack_info *ctinfo,
572  u_int8_t pf, unsigned int hooknum)
573 {
574  struct dccp_hdr _dh, *dh;
575  unsigned int dccp_len = skb->len - dataoff;
576  unsigned int cscov;
577  const char *msg;
578 
579  dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
580  if (dh == NULL) {
581  msg = "nf_ct_dccp: short packet ";
582  goto out_invalid;
583  }
584 
585  if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
586  dh->dccph_doff * 4 > dccp_len) {
587  msg = "nf_ct_dccp: truncated/malformed packet ";
588  goto out_invalid;
589  }
590 
591  cscov = dccp_len;
592  if (dh->dccph_cscov) {
593  cscov = (dh->dccph_cscov - 1) * 4;
594  if (cscov > dccp_len) {
595  msg = "nf_ct_dccp: bad checksum coverage ";
596  goto out_invalid;
597  }
598  }
599 
600  if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
601  nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP,
602  pf)) {
603  msg = "nf_ct_dccp: bad checksum ";
604  goto out_invalid;
605  }
606 
607  if (dh->dccph_type >= DCCP_PKT_INVALID) {
608  msg = "nf_ct_dccp: reserved packet type ";
609  goto out_invalid;
610  }
611 
612  return NF_ACCEPT;
613 
614 out_invalid:
615  if (LOG_INVALID(net, IPPROTO_DCCP))
616  nf_log_packet(pf, 0, skb, NULL, NULL, NULL, msg);
617  return -NF_ACCEPT;
618 }
619 
620 static int dccp_print_tuple(struct seq_file *s,
621  const struct nf_conntrack_tuple *tuple)
622 {
623  return seq_printf(s, "sport=%hu dport=%hu ",
624  ntohs(tuple->src.u.dccp.port),
625  ntohs(tuple->dst.u.dccp.port));
626 }
627 
628 static int dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
629 {
630  return seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
631 }
632 
633 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
634 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
635  struct nf_conn *ct)
636 {
637  struct nlattr *nest_parms;
638 
639  spin_lock_bh(&ct->lock);
640  nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED);
641  if (!nest_parms)
642  goto nla_put_failure;
643  if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state) ||
644  nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE,
645  ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) ||
646  nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ,
647  cpu_to_be64(ct->proto.dccp.handshake_seq)))
648  goto nla_put_failure;
649  nla_nest_end(skb, nest_parms);
650  spin_unlock_bh(&ct->lock);
651  return 0;
652 
653 nla_put_failure:
654  spin_unlock_bh(&ct->lock);
655  return -1;
656 }
657 
658 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
660  [CTA_PROTOINFO_DCCP_ROLE] = { .type = NLA_U8 },
662 };
663 
664 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
665 {
666  struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
667  struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
668  int err;
669 
670  if (!attr)
671  return 0;
672 
673  err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr,
674  dccp_nla_policy);
675  if (err < 0)
676  return err;
677 
678  if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
680  nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX ||
681  nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) {
682  return -EINVAL;
683  }
684 
685  spin_lock_bh(&ct->lock);
686  ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
687  if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) {
690  } else {
693  }
695  ct->proto.dccp.handshake_seq =
696  be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]));
697  }
698  spin_unlock_bh(&ct->lock);
699  return 0;
700 }
701 
702 static int dccp_nlattr_size(void)
703 {
704  return nla_total_size(0) /* CTA_PROTOINFO_DCCP */
705  + nla_policy_len(dccp_nla_policy, CTA_PROTOINFO_DCCP_MAX + 1);
706 }
707 
708 #endif
709 
710 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
711 
712 #include <linux/netfilter/nfnetlink.h>
714 
715 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[],
716  struct net *net, void *data)
717 {
718  struct dccp_net *dn = dccp_pernet(net);
719  unsigned int *timeouts = data;
720  int i;
721 
722  /* set default DCCP timeouts. */
723  for (i=0; i<CT_DCCP_MAX; i++)
724  timeouts[i] = dn->dccp_timeout[i];
725 
726  /* there's a 1:1 mapping between attributes and protocol states. */
728  if (tb[i]) {
729  timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ;
730  }
731  }
732  return 0;
733 }
734 
735 static int
736 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
737 {
738  const unsigned int *timeouts = data;
739  int i;
740 
741  for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
742  if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ)))
743  goto nla_put_failure;
744  }
745  return 0;
746 
747 nla_put_failure:
748  return -ENOSPC;
749 }
750 
751 static const struct nla_policy
752 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = {
754  [CTA_TIMEOUT_DCCP_RESPOND] = { .type = NLA_U32 },
755  [CTA_TIMEOUT_DCCP_PARTOPEN] = { .type = NLA_U32 },
756  [CTA_TIMEOUT_DCCP_OPEN] = { .type = NLA_U32 },
757  [CTA_TIMEOUT_DCCP_CLOSEREQ] = { .type = NLA_U32 },
758  [CTA_TIMEOUT_DCCP_CLOSING] = { .type = NLA_U32 },
759  [CTA_TIMEOUT_DCCP_TIMEWAIT] = { .type = NLA_U32 },
760 };
761 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
762 
763 #ifdef CONFIG_SYSCTL
764 /* template, data assigned later */
765 static struct ctl_table dccp_sysctl_table[] = {
766  {
767  .procname = "nf_conntrack_dccp_timeout_request",
768  .maxlen = sizeof(unsigned int),
769  .mode = 0644,
771  },
772  {
773  .procname = "nf_conntrack_dccp_timeout_respond",
774  .maxlen = sizeof(unsigned int),
775  .mode = 0644,
777  },
778  {
779  .procname = "nf_conntrack_dccp_timeout_partopen",
780  .maxlen = sizeof(unsigned int),
781  .mode = 0644,
783  },
784  {
785  .procname = "nf_conntrack_dccp_timeout_open",
786  .maxlen = sizeof(unsigned int),
787  .mode = 0644,
789  },
790  {
791  .procname = "nf_conntrack_dccp_timeout_closereq",
792  .maxlen = sizeof(unsigned int),
793  .mode = 0644,
795  },
796  {
797  .procname = "nf_conntrack_dccp_timeout_closing",
798  .maxlen = sizeof(unsigned int),
799  .mode = 0644,
801  },
802  {
803  .procname = "nf_conntrack_dccp_timeout_timewait",
804  .maxlen = sizeof(unsigned int),
805  .mode = 0644,
807  },
808  {
809  .procname = "nf_conntrack_dccp_loose",
810  .maxlen = sizeof(int),
811  .mode = 0644,
813  },
814  { }
815 };
816 #endif /* CONFIG_SYSCTL */
817 
818 static int dccp_kmemdup_sysctl_table(struct nf_proto_net *pn,
819  struct dccp_net *dn)
820 {
821 #ifdef CONFIG_SYSCTL
822  if (pn->ctl_table)
823  return 0;
824 
825  pn->ctl_table = kmemdup(dccp_sysctl_table,
826  sizeof(dccp_sysctl_table),
827  GFP_KERNEL);
828  if (!pn->ctl_table)
829  return -ENOMEM;
830 
831  pn->ctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST];
832  pn->ctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND];
833  pn->ctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN];
834  pn->ctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN];
835  pn->ctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ];
836  pn->ctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING];
837  pn->ctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT];
838  pn->ctl_table[7].data = &dn->dccp_loose;
839 #endif
840  return 0;
841 }
842 
843 static int dccp_init_net(struct net *net, u_int16_t proto)
844 {
845  struct dccp_net *dn = dccp_pernet(net);
846  struct nf_proto_net *pn = &dn->pn;
847 
848  if (!pn->users) {
849  /* default values */
850  dn->dccp_loose = 1;
854  dn->dccp_timeout[CT_DCCP_OPEN] = 12 * 3600 * HZ;
855  dn->dccp_timeout[CT_DCCP_CLOSEREQ] = 64 * HZ;
856  dn->dccp_timeout[CT_DCCP_CLOSING] = 64 * HZ;
858  }
859 
860  return dccp_kmemdup_sysctl_table(pn, dn);
861 }
862 
863 static struct nf_conntrack_l4proto dccp_proto4 __read_mostly = {
864  .l3proto = AF_INET,
865  .l4proto = IPPROTO_DCCP,
866  .name = "dccp",
867  .pkt_to_tuple = dccp_pkt_to_tuple,
868  .invert_tuple = dccp_invert_tuple,
869  .new = dccp_new,
870  .packet = dccp_packet,
871  .get_timeouts = dccp_get_timeouts,
872  .error = dccp_error,
873  .print_tuple = dccp_print_tuple,
874  .print_conntrack = dccp_print_conntrack,
875 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
876  .to_nlattr = dccp_to_nlattr,
877  .nlattr_size = dccp_nlattr_size,
878  .from_nlattr = nlattr_to_dccp,
879  .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
880  .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
881  .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
882  .nla_policy = nf_ct_port_nla_policy,
883 #endif
884 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
885  .ctnl_timeout = {
886  .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
887  .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
888  .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
889  .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
890  .nla_policy = dccp_timeout_nla_policy,
891  },
892 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
893  .net_id = &dccp_net_id,
894  .init_net = dccp_init_net,
895 };
896 
897 static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = {
898  .l3proto = AF_INET6,
899  .l4proto = IPPROTO_DCCP,
900  .name = "dccp",
901  .pkt_to_tuple = dccp_pkt_to_tuple,
902  .invert_tuple = dccp_invert_tuple,
903  .new = dccp_new,
904  .packet = dccp_packet,
905  .get_timeouts = dccp_get_timeouts,
906  .error = dccp_error,
907  .print_tuple = dccp_print_tuple,
908  .print_conntrack = dccp_print_conntrack,
909 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
910  .to_nlattr = dccp_to_nlattr,
911  .nlattr_size = dccp_nlattr_size,
912  .from_nlattr = nlattr_to_dccp,
913  .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
914  .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
915  .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
916  .nla_policy = nf_ct_port_nla_policy,
917 #endif
918 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
919  .ctnl_timeout = {
920  .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
921  .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
922  .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
923  .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
924  .nla_policy = dccp_timeout_nla_policy,
925  },
926 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
927  .net_id = &dccp_net_id,
928  .init_net = dccp_init_net,
929 };
930 
931 static __net_init int dccp_net_init(struct net *net)
932 {
933  int ret = 0;
935  &dccp_proto4);
936  if (ret < 0) {
937  pr_err("nf_conntrack_l4proto_dccp4 :protocol register failed.\n");
938  goto out;
939  }
941  &dccp_proto6);
942  if (ret < 0) {
943  pr_err("nf_conntrack_l4proto_dccp6 :protocol register failed.\n");
944  goto cleanup_dccp4;
945  }
946  return 0;
947 cleanup_dccp4:
949  &dccp_proto4);
950 out:
951  return ret;
952 }
953 
954 static __net_exit void dccp_net_exit(struct net *net)
955 {
957  &dccp_proto6);
959  &dccp_proto4);
960 }
961 
962 static struct pernet_operations dccp_net_ops = {
963  .init = dccp_net_init,
964  .exit = dccp_net_exit,
965  .id = &dccp_net_id,
966  .size = sizeof(struct dccp_net),
967 };
968 
969 static int __init nf_conntrack_proto_dccp_init(void)
970 {
971  return register_pernet_subsys(&dccp_net_ops);
972 }
973 
974 static void __exit nf_conntrack_proto_dccp_fini(void)
975 {
976  unregister_pernet_subsys(&dccp_net_ops);
977 }
978 
979 module_init(nf_conntrack_proto_dccp_init);
980 module_exit(nf_conntrack_proto_dccp_fini);
981 
982 MODULE_AUTHOR("Patrick McHardy <[email protected]>");
983 MODULE_DESCRIPTION("DCCP connection tracking protocol helper");
984 MODULE_LICENSE("GPL");