netfilter: invoke synchronize_rcu after set the _hook_ to NULL
[GitHub/mt8127/android_kernel_alcatel_ttab.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 #include <linux/slab.h>
19
20 #include <net/net_namespace.h>
21 #include <net/netns/generic.h>
22
23 #include <linux/netfilter/nfnetlink_conntrack.h>
24 #include <net/netfilter/nf_conntrack.h>
25 #include <net/netfilter/nf_conntrack_l4proto.h>
26 #include <net/netfilter/nf_conntrack_ecache.h>
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;
391 int dccp_loose;
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
449 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
450 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
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(net, nf_ct_l3num(ct), 0, skb, NULL, NULL,
460 NULL, msg);
461 return false;
462 }
463
464 static u64 dccp_ack_seq(const struct dccp_hdr *dh)
465 {
466 const struct dccp_hdr_ack_bits *dhack;
467
468 dhack = (void *)dh + __dccp_basic_hdr_len(dh);
469 return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) +
470 ntohl(dhack->dccph_ack_nr_low);
471 }
472
473 static unsigned int *dccp_get_timeouts(struct net *net)
474 {
475 return dccp_pernet(net)->dccp_timeout;
476 }
477
478 static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
479 unsigned int dataoff, enum ip_conntrack_info ctinfo,
480 u_int8_t pf, unsigned int hooknum,
481 unsigned int *timeouts)
482 {
483 struct net *net = nf_ct_net(ct);
484 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
485 struct dccp_hdr _dh, *dh;
486 u_int8_t type, old_state, new_state;
487 enum ct_dccp_roles role;
488
489 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh);
490 BUG_ON(dh == NULL);
491 type = dh->dccph_type;
492
493 if (type == DCCP_PKT_RESET &&
494 !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
495 /* Tear down connection immediately if only reply is a RESET */
496 nf_ct_kill_acct(ct, ctinfo, skb);
497 return NF_ACCEPT;
498 }
499
500 spin_lock_bh(&ct->lock);
501
502 role = ct->proto.dccp.role[dir];
503 old_state = ct->proto.dccp.state;
504 new_state = dccp_state_table[role][type][old_state];
505
506 switch (new_state) {
507 case CT_DCCP_REQUEST:
508 if (old_state == CT_DCCP_TIMEWAIT &&
509 role == CT_DCCP_ROLE_SERVER) {
510 /* Reincarnation in the reverse direction: reopen and
511 * reverse client/server roles. */
512 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
513 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
514 }
515 break;
516 case CT_DCCP_RESPOND:
517 if (old_state == CT_DCCP_REQUEST)
518 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
519 break;
520 case CT_DCCP_PARTOPEN:
521 if (old_state == CT_DCCP_RESPOND &&
522 type == DCCP_PKT_ACK &&
523 dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
524 set_bit(IPS_ASSURED_BIT, &ct->status);
525 break;
526 case CT_DCCP_IGNORE:
527 /*
528 * Connection tracking might be out of sync, so we ignore
529 * packets that might establish a new connection and resync
530 * if the server responds with a valid Response.
531 */
532 if (ct->proto.dccp.last_dir == !dir &&
533 ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
534 type == DCCP_PKT_RESPONSE) {
535 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
536 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
537 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
538 new_state = CT_DCCP_RESPOND;
539 break;
540 }
541 ct->proto.dccp.last_dir = dir;
542 ct->proto.dccp.last_pkt = type;
543
544 spin_unlock_bh(&ct->lock);
545 if (LOG_INVALID(net, IPPROTO_DCCP))
546 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
547 "nf_ct_dccp: invalid packet ignored ");
548 return NF_ACCEPT;
549 case CT_DCCP_INVALID:
550 spin_unlock_bh(&ct->lock);
551 if (LOG_INVALID(net, IPPROTO_DCCP))
552 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
553 "nf_ct_dccp: invalid state transition ");
554 return -NF_ACCEPT;
555 }
556
557 ct->proto.dccp.last_dir = dir;
558 ct->proto.dccp.last_pkt = type;
559 ct->proto.dccp.state = new_state;
560 spin_unlock_bh(&ct->lock);
561
562 if (new_state != old_state)
563 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
564
565 nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]);
566
567 return NF_ACCEPT;
568 }
569
570 static int dccp_error(struct net *net, struct nf_conn *tmpl,
571 struct sk_buff *skb, unsigned int dataoff,
572 enum ip_conntrack_info *ctinfo,
573 u_int8_t pf, unsigned int hooknum)
574 {
575 struct dccp_hdr _dh, *dh;
576 unsigned int dccp_len = skb->len - dataoff;
577 unsigned int cscov;
578 const char *msg;
579
580 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh);
581 if (dh == NULL) {
582 msg = "nf_ct_dccp: short packet ";
583 goto out_invalid;
584 }
585
586 if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
587 dh->dccph_doff * 4 > dccp_len) {
588 msg = "nf_ct_dccp: truncated/malformed packet ";
589 goto out_invalid;
590 }
591
592 cscov = dccp_len;
593 if (dh->dccph_cscov) {
594 cscov = (dh->dccph_cscov - 1) * 4;
595 if (cscov > dccp_len) {
596 msg = "nf_ct_dccp: bad checksum coverage ";
597 goto out_invalid;
598 }
599 }
600
601 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
602 nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP,
603 pf)) {
604 msg = "nf_ct_dccp: bad checksum ";
605 goto out_invalid;
606 }
607
608 if (dh->dccph_type >= DCCP_PKT_INVALID) {
609 msg = "nf_ct_dccp: reserved packet type ";
610 goto out_invalid;
611 }
612
613 return NF_ACCEPT;
614
615 out_invalid:
616 if (LOG_INVALID(net, IPPROTO_DCCP))
617 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL, msg);
618 return -NF_ACCEPT;
619 }
620
621 static int dccp_print_tuple(struct seq_file *s,
622 const struct nf_conntrack_tuple *tuple)
623 {
624 return seq_printf(s, "sport=%hu dport=%hu ",
625 ntohs(tuple->src.u.dccp.port),
626 ntohs(tuple->dst.u.dccp.port));
627 }
628
629 static int dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
630 {
631 return seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
632 }
633
634 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
635 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
636 struct nf_conn *ct)
637 {
638 struct nlattr *nest_parms;
639
640 spin_lock_bh(&ct->lock);
641 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED);
642 if (!nest_parms)
643 goto nla_put_failure;
644 if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state) ||
645 nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE,
646 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) ||
647 nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ,
648 cpu_to_be64(ct->proto.dccp.handshake_seq)))
649 goto nla_put_failure;
650 nla_nest_end(skb, nest_parms);
651 spin_unlock_bh(&ct->lock);
652 return 0;
653
654 nla_put_failure:
655 spin_unlock_bh(&ct->lock);
656 return -1;
657 }
658
659 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
660 [CTA_PROTOINFO_DCCP_STATE] = { .type = NLA_U8 },
661 [CTA_PROTOINFO_DCCP_ROLE] = { .type = NLA_U8 },
662 [CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ] = { .type = NLA_U64 },
663 };
664
665 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
666 {
667 struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
668 struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
669 int err;
670
671 if (!attr)
672 return 0;
673
674 err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr,
675 dccp_nla_policy);
676 if (err < 0)
677 return err;
678
679 if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
680 !tb[CTA_PROTOINFO_DCCP_ROLE] ||
681 nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX ||
682 nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) {
683 return -EINVAL;
684 }
685
686 spin_lock_bh(&ct->lock);
687 ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
688 if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) {
689 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
690 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
691 } else {
692 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_SERVER;
693 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_CLIENT;
694 }
695 if (tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]) {
696 ct->proto.dccp.handshake_seq =
697 be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]));
698 }
699 spin_unlock_bh(&ct->lock);
700 return 0;
701 }
702
703 static int dccp_nlattr_size(void)
704 {
705 return nla_total_size(0) /* CTA_PROTOINFO_DCCP */
706 + nla_policy_len(dccp_nla_policy, CTA_PROTOINFO_DCCP_MAX + 1);
707 }
708
709 #endif
710
711 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
712
713 #include <linux/netfilter/nfnetlink.h>
714 #include <linux/netfilter/nfnetlink_cttimeout.h>
715
716 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[],
717 struct net *net, void *data)
718 {
719 struct dccp_net *dn = dccp_pernet(net);
720 unsigned int *timeouts = data;
721 int i;
722
723 /* set default DCCP timeouts. */
724 for (i=0; i<CT_DCCP_MAX; i++)
725 timeouts[i] = dn->dccp_timeout[i];
726
727 /* there's a 1:1 mapping between attributes and protocol states. */
728 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
729 if (tb[i]) {
730 timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ;
731 }
732 }
733 return 0;
734 }
735
736 static int
737 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
738 {
739 const unsigned int *timeouts = data;
740 int i;
741
742 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
743 if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ)))
744 goto nla_put_failure;
745 }
746 return 0;
747
748 nla_put_failure:
749 return -ENOSPC;
750 }
751
752 static const struct nla_policy
753 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = {
754 [CTA_TIMEOUT_DCCP_REQUEST] = { .type = NLA_U32 },
755 [CTA_TIMEOUT_DCCP_RESPOND] = { .type = NLA_U32 },
756 [CTA_TIMEOUT_DCCP_PARTOPEN] = { .type = NLA_U32 },
757 [CTA_TIMEOUT_DCCP_OPEN] = { .type = NLA_U32 },
758 [CTA_TIMEOUT_DCCP_CLOSEREQ] = { .type = NLA_U32 },
759 [CTA_TIMEOUT_DCCP_CLOSING] = { .type = NLA_U32 },
760 [CTA_TIMEOUT_DCCP_TIMEWAIT] = { .type = NLA_U32 },
761 };
762 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
763
764 #ifdef CONFIG_SYSCTL
765 /* template, data assigned later */
766 static struct ctl_table dccp_sysctl_table[] = {
767 {
768 .procname = "nf_conntrack_dccp_timeout_request",
769 .maxlen = sizeof(unsigned int),
770 .mode = 0644,
771 .proc_handler = proc_dointvec_jiffies,
772 },
773 {
774 .procname = "nf_conntrack_dccp_timeout_respond",
775 .maxlen = sizeof(unsigned int),
776 .mode = 0644,
777 .proc_handler = proc_dointvec_jiffies,
778 },
779 {
780 .procname = "nf_conntrack_dccp_timeout_partopen",
781 .maxlen = sizeof(unsigned int),
782 .mode = 0644,
783 .proc_handler = proc_dointvec_jiffies,
784 },
785 {
786 .procname = "nf_conntrack_dccp_timeout_open",
787 .maxlen = sizeof(unsigned int),
788 .mode = 0644,
789 .proc_handler = proc_dointvec_jiffies,
790 },
791 {
792 .procname = "nf_conntrack_dccp_timeout_closereq",
793 .maxlen = sizeof(unsigned int),
794 .mode = 0644,
795 .proc_handler = proc_dointvec_jiffies,
796 },
797 {
798 .procname = "nf_conntrack_dccp_timeout_closing",
799 .maxlen = sizeof(unsigned int),
800 .mode = 0644,
801 .proc_handler = proc_dointvec_jiffies,
802 },
803 {
804 .procname = "nf_conntrack_dccp_timeout_timewait",
805 .maxlen = sizeof(unsigned int),
806 .mode = 0644,
807 .proc_handler = proc_dointvec_jiffies,
808 },
809 {
810 .procname = "nf_conntrack_dccp_loose",
811 .maxlen = sizeof(int),
812 .mode = 0644,
813 .proc_handler = proc_dointvec,
814 },
815 { }
816 };
817 #endif /* CONFIG_SYSCTL */
818
819 static int dccp_kmemdup_sysctl_table(struct net *net, struct nf_proto_net *pn,
820 struct dccp_net *dn)
821 {
822 #ifdef CONFIG_SYSCTL
823 if (pn->ctl_table)
824 return 0;
825
826 pn->ctl_table = kmemdup(dccp_sysctl_table,
827 sizeof(dccp_sysctl_table),
828 GFP_KERNEL);
829 if (!pn->ctl_table)
830 return -ENOMEM;
831
832 pn->ctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST];
833 pn->ctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND];
834 pn->ctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN];
835 pn->ctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN];
836 pn->ctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ];
837 pn->ctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING];
838 pn->ctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT];
839 pn->ctl_table[7].data = &dn->dccp_loose;
840
841 /* Don't export sysctls to unprivileged users */
842 if (net->user_ns != &init_user_ns)
843 pn->ctl_table[0].procname = NULL;
844 #endif
845 return 0;
846 }
847
848 static int dccp_init_net(struct net *net, u_int16_t proto)
849 {
850 struct dccp_net *dn = dccp_pernet(net);
851 struct nf_proto_net *pn = &dn->pn;
852
853 if (!pn->users) {
854 /* default values */
855 dn->dccp_loose = 1;
856 dn->dccp_timeout[CT_DCCP_REQUEST] = 2 * DCCP_MSL;
857 dn->dccp_timeout[CT_DCCP_RESPOND] = 4 * DCCP_MSL;
858 dn->dccp_timeout[CT_DCCP_PARTOPEN] = 4 * DCCP_MSL;
859 dn->dccp_timeout[CT_DCCP_OPEN] = 12 * 3600 * HZ;
860 dn->dccp_timeout[CT_DCCP_CLOSEREQ] = 64 * HZ;
861 dn->dccp_timeout[CT_DCCP_CLOSING] = 64 * HZ;
862 dn->dccp_timeout[CT_DCCP_TIMEWAIT] = 2 * DCCP_MSL;
863 }
864
865 return dccp_kmemdup_sysctl_table(net, pn, dn);
866 }
867
868 static struct nf_conntrack_l4proto dccp_proto4 __read_mostly = {
869 .l3proto = AF_INET,
870 .l4proto = IPPROTO_DCCP,
871 .name = "dccp",
872 .pkt_to_tuple = dccp_pkt_to_tuple,
873 .invert_tuple = dccp_invert_tuple,
874 .new = dccp_new,
875 .packet = dccp_packet,
876 .get_timeouts = dccp_get_timeouts,
877 .error = dccp_error,
878 .print_tuple = dccp_print_tuple,
879 .print_conntrack = dccp_print_conntrack,
880 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
881 .to_nlattr = dccp_to_nlattr,
882 .nlattr_size = dccp_nlattr_size,
883 .from_nlattr = nlattr_to_dccp,
884 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
885 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
886 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
887 .nla_policy = nf_ct_port_nla_policy,
888 #endif
889 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
890 .ctnl_timeout = {
891 .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
892 .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
893 .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
894 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
895 .nla_policy = dccp_timeout_nla_policy,
896 },
897 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
898 .net_id = &dccp_net_id,
899 .init_net = dccp_init_net,
900 };
901
902 static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = {
903 .l3proto = AF_INET6,
904 .l4proto = IPPROTO_DCCP,
905 .name = "dccp",
906 .pkt_to_tuple = dccp_pkt_to_tuple,
907 .invert_tuple = dccp_invert_tuple,
908 .new = dccp_new,
909 .packet = dccp_packet,
910 .get_timeouts = dccp_get_timeouts,
911 .error = dccp_error,
912 .print_tuple = dccp_print_tuple,
913 .print_conntrack = dccp_print_conntrack,
914 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
915 .to_nlattr = dccp_to_nlattr,
916 .nlattr_size = dccp_nlattr_size,
917 .from_nlattr = nlattr_to_dccp,
918 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
919 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
920 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
921 .nla_policy = nf_ct_port_nla_policy,
922 #endif
923 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
924 .ctnl_timeout = {
925 .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
926 .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
927 .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
928 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
929 .nla_policy = dccp_timeout_nla_policy,
930 },
931 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
932 .net_id = &dccp_net_id,
933 .init_net = dccp_init_net,
934 };
935
936 static __net_init int dccp_net_init(struct net *net)
937 {
938 int ret = 0;
939 ret = nf_ct_l4proto_pernet_register(net, &dccp_proto4);
940 if (ret < 0) {
941 pr_err("nf_conntrack_dccp4: pernet registration failed.\n");
942 goto out;
943 }
944 ret = nf_ct_l4proto_pernet_register(net, &dccp_proto6);
945 if (ret < 0) {
946 pr_err("nf_conntrack_dccp6: pernet registration failed.\n");
947 goto cleanup_dccp4;
948 }
949 return 0;
950 cleanup_dccp4:
951 nf_ct_l4proto_pernet_unregister(net, &dccp_proto4);
952 out:
953 return ret;
954 }
955
956 static __net_exit void dccp_net_exit(struct net *net)
957 {
958 nf_ct_l4proto_pernet_unregister(net, &dccp_proto6);
959 nf_ct_l4proto_pernet_unregister(net, &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 int ret;
972
973 ret = register_pernet_subsys(&dccp_net_ops);
974 if (ret < 0)
975 goto out_pernet;
976
977 ret = nf_ct_l4proto_register(&dccp_proto4);
978 if (ret < 0)
979 goto out_dccp4;
980
981 ret = nf_ct_l4proto_register(&dccp_proto6);
982 if (ret < 0)
983 goto out_dccp6;
984
985 return 0;
986 out_dccp6:
987 nf_ct_l4proto_unregister(&dccp_proto4);
988 out_dccp4:
989 unregister_pernet_subsys(&dccp_net_ops);
990 out_pernet:
991 return ret;
992 }
993
994 static void __exit nf_conntrack_proto_dccp_fini(void)
995 {
996 nf_ct_l4proto_unregister(&dccp_proto6);
997 nf_ct_l4proto_unregister(&dccp_proto4);
998 unregister_pernet_subsys(&dccp_net_ops);
999 }
1000
1001 module_init(nf_conntrack_proto_dccp_init);
1002 module_exit(nf_conntrack_proto_dccp_fini);
1003
1004 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1005 MODULE_DESCRIPTION("DCCP connection tracking protocol helper");
1006 MODULE_LICENSE("GPL");