]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - net/netfilter/nf_conntrack_proto_dccp.c
netfilter: conntrack: handle builtin l4proto packet functions via direct calls
[thirdparty/kernel/stable.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/init.h>
13 #include <linux/sysctl.h>
14 #include <linux/spinlock.h>
15 #include <linux/skbuff.h>
16 #include <linux/dccp.h>
17 #include <linux/slab.h>
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_conntrack_ecache.h>
26 #include <net/netfilter/nf_conntrack_timeout.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 sIV, 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 sIV, 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 sIV, 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 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
383 },
384 },
385 };
386
387 static noinline bool
388 dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
389 const struct dccp_hdr *dh)
390 {
391 struct net *net = nf_ct_net(ct);
392 struct nf_dccp_net *dn;
393 const char *msg;
394 u_int8_t state;
395
396 state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
397 switch (state) {
398 default:
399 dn = nf_dccp_pernet(net);
400 if (dn->dccp_loose == 0) {
401 msg = "not picking up existing connection ";
402 goto out_invalid;
403 }
404 case CT_DCCP_REQUEST:
405 break;
406 case CT_DCCP_INVALID:
407 msg = "invalid state transition ";
408 goto out_invalid;
409 }
410
411 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
412 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
413 ct->proto.dccp.state = CT_DCCP_NONE;
414 ct->proto.dccp.last_pkt = DCCP_PKT_REQUEST;
415 ct->proto.dccp.last_dir = IP_CT_DIR_ORIGINAL;
416 ct->proto.dccp.handshake_seq = 0;
417 return true;
418
419 out_invalid:
420 nf_ct_l4proto_log_invalid(skb, ct, "%s", msg);
421 return false;
422 }
423
424 static u64 dccp_ack_seq(const struct dccp_hdr *dh)
425 {
426 const struct dccp_hdr_ack_bits *dhack;
427
428 dhack = (void *)dh + __dccp_basic_hdr_len(dh);
429 return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) +
430 ntohl(dhack->dccph_ack_nr_low);
431 }
432
433 static bool dccp_error(const struct dccp_hdr *dh,
434 struct sk_buff *skb, unsigned int dataoff,
435 const struct nf_hook_state *state)
436 {
437 unsigned int dccp_len = skb->len - dataoff;
438 unsigned int cscov;
439 const char *msg;
440
441 if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
442 dh->dccph_doff * 4 > dccp_len) {
443 msg = "nf_ct_dccp: truncated/malformed packet ";
444 goto out_invalid;
445 }
446
447 cscov = dccp_len;
448 if (dh->dccph_cscov) {
449 cscov = (dh->dccph_cscov - 1) * 4;
450 if (cscov > dccp_len) {
451 msg = "nf_ct_dccp: bad checksum coverage ";
452 goto out_invalid;
453 }
454 }
455
456 if (state->hook == NF_INET_PRE_ROUTING &&
457 state->net->ct.sysctl_checksum &&
458 nf_checksum_partial(skb, state->hook, dataoff, cscov,
459 IPPROTO_DCCP, state->pf)) {
460 msg = "nf_ct_dccp: bad checksum ";
461 goto out_invalid;
462 }
463
464 if (dh->dccph_type >= DCCP_PKT_INVALID) {
465 msg = "nf_ct_dccp: reserved packet type ";
466 goto out_invalid;
467 }
468 return false;
469 out_invalid:
470 nf_l4proto_log_invalid(skb, state->net, state->pf,
471 IPPROTO_DCCP, "%s", msg);
472 return true;
473 }
474
475 int nf_conntrack_dccp_packet(struct nf_conn *ct, struct sk_buff *skb,
476 unsigned int dataoff,
477 enum ip_conntrack_info ctinfo,
478 const struct nf_hook_state *state)
479 {
480 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
481 struct dccp_hdr _dh, *dh;
482 u_int8_t type, old_state, new_state;
483 enum ct_dccp_roles role;
484 unsigned int *timeouts;
485
486 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh);
487 if (!dh)
488 return NF_DROP;
489
490 if (dccp_error(dh, skb, dataoff, state))
491 return -NF_ACCEPT;
492
493 type = dh->dccph_type;
494 if (!nf_ct_is_confirmed(ct) && !dccp_new(ct, skb, dh))
495 return -NF_ACCEPT;
496
497 if (type == DCCP_PKT_RESET &&
498 !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
499 /* Tear down connection immediately if only reply is a RESET */
500 nf_ct_kill_acct(ct, ctinfo, skb);
501 return NF_ACCEPT;
502 }
503
504 spin_lock_bh(&ct->lock);
505
506 role = ct->proto.dccp.role[dir];
507 old_state = ct->proto.dccp.state;
508 new_state = dccp_state_table[role][type][old_state];
509
510 switch (new_state) {
511 case CT_DCCP_REQUEST:
512 if (old_state == CT_DCCP_TIMEWAIT &&
513 role == CT_DCCP_ROLE_SERVER) {
514 /* Reincarnation in the reverse direction: reopen and
515 * reverse client/server roles. */
516 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
517 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
518 }
519 break;
520 case CT_DCCP_RESPOND:
521 if (old_state == CT_DCCP_REQUEST)
522 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
523 break;
524 case CT_DCCP_PARTOPEN:
525 if (old_state == CT_DCCP_RESPOND &&
526 type == DCCP_PKT_ACK &&
527 dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
528 set_bit(IPS_ASSURED_BIT, &ct->status);
529 break;
530 case CT_DCCP_IGNORE:
531 /*
532 * Connection tracking might be out of sync, so we ignore
533 * packets that might establish a new connection and resync
534 * if the server responds with a valid Response.
535 */
536 if (ct->proto.dccp.last_dir == !dir &&
537 ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
538 type == DCCP_PKT_RESPONSE) {
539 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
540 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
541 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
542 new_state = CT_DCCP_RESPOND;
543 break;
544 }
545 ct->proto.dccp.last_dir = dir;
546 ct->proto.dccp.last_pkt = type;
547
548 spin_unlock_bh(&ct->lock);
549 nf_ct_l4proto_log_invalid(skb, ct, "%s", "invalid packet");
550 return NF_ACCEPT;
551 case CT_DCCP_INVALID:
552 spin_unlock_bh(&ct->lock);
553 nf_ct_l4proto_log_invalid(skb, ct, "%s", "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 timeouts = nf_ct_timeout_lookup(ct);
566 if (!timeouts)
567 timeouts = nf_dccp_pernet(nf_ct_net(ct))->dccp_timeout;
568 nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]);
569
570 return NF_ACCEPT;
571 }
572
573 static bool dccp_can_early_drop(const struct nf_conn *ct)
574 {
575 switch (ct->proto.dccp.state) {
576 case CT_DCCP_CLOSEREQ:
577 case CT_DCCP_CLOSING:
578 case CT_DCCP_TIMEWAIT:
579 return true;
580 default:
581 break;
582 }
583
584 return false;
585 }
586
587 #ifdef CONFIG_NF_CONNTRACK_PROCFS
588 static void dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
589 {
590 seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
591 }
592 #endif
593
594 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
595 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
596 struct nf_conn *ct)
597 {
598 struct nlattr *nest_parms;
599
600 spin_lock_bh(&ct->lock);
601 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED);
602 if (!nest_parms)
603 goto nla_put_failure;
604 if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state) ||
605 nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE,
606 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) ||
607 nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ,
608 cpu_to_be64(ct->proto.dccp.handshake_seq),
609 CTA_PROTOINFO_DCCP_PAD))
610 goto nla_put_failure;
611 nla_nest_end(skb, nest_parms);
612 spin_unlock_bh(&ct->lock);
613 return 0;
614
615 nla_put_failure:
616 spin_unlock_bh(&ct->lock);
617 return -1;
618 }
619
620 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
621 [CTA_PROTOINFO_DCCP_STATE] = { .type = NLA_U8 },
622 [CTA_PROTOINFO_DCCP_ROLE] = { .type = NLA_U8 },
623 [CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ] = { .type = NLA_U64 },
624 [CTA_PROTOINFO_DCCP_PAD] = { .type = NLA_UNSPEC },
625 };
626
627 #define DCCP_NLATTR_SIZE ( \
628 NLA_ALIGN(NLA_HDRLEN + 1) + \
629 NLA_ALIGN(NLA_HDRLEN + 1) + \
630 NLA_ALIGN(NLA_HDRLEN + sizeof(u64)) + \
631 NLA_ALIGN(NLA_HDRLEN + 0))
632
633 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
634 {
635 struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
636 struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
637 int err;
638
639 if (!attr)
640 return 0;
641
642 err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr,
643 dccp_nla_policy, NULL);
644 if (err < 0)
645 return err;
646
647 if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
648 !tb[CTA_PROTOINFO_DCCP_ROLE] ||
649 nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX ||
650 nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) {
651 return -EINVAL;
652 }
653
654 spin_lock_bh(&ct->lock);
655 ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
656 if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) {
657 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
658 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
659 } else {
660 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_SERVER;
661 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_CLIENT;
662 }
663 if (tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]) {
664 ct->proto.dccp.handshake_seq =
665 be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]));
666 }
667 spin_unlock_bh(&ct->lock);
668 return 0;
669 }
670 #endif
671
672 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
673
674 #include <linux/netfilter/nfnetlink.h>
675 #include <linux/netfilter/nfnetlink_cttimeout.h>
676
677 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[],
678 struct net *net, void *data)
679 {
680 struct nf_dccp_net *dn = nf_dccp_pernet(net);
681 unsigned int *timeouts = data;
682 int i;
683
684 /* set default DCCP timeouts. */
685 for (i=0; i<CT_DCCP_MAX; i++)
686 timeouts[i] = dn->dccp_timeout[i];
687
688 /* there's a 1:1 mapping between attributes and protocol states. */
689 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
690 if (tb[i]) {
691 timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ;
692 }
693 }
694
695 timeouts[CTA_TIMEOUT_DCCP_UNSPEC] = timeouts[CTA_TIMEOUT_DCCP_REQUEST];
696 return 0;
697 }
698
699 static int
700 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
701 {
702 const unsigned int *timeouts = data;
703 int i;
704
705 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
706 if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ)))
707 goto nla_put_failure;
708 }
709 return 0;
710
711 nla_put_failure:
712 return -ENOSPC;
713 }
714
715 static const struct nla_policy
716 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = {
717 [CTA_TIMEOUT_DCCP_REQUEST] = { .type = NLA_U32 },
718 [CTA_TIMEOUT_DCCP_RESPOND] = { .type = NLA_U32 },
719 [CTA_TIMEOUT_DCCP_PARTOPEN] = { .type = NLA_U32 },
720 [CTA_TIMEOUT_DCCP_OPEN] = { .type = NLA_U32 },
721 [CTA_TIMEOUT_DCCP_CLOSEREQ] = { .type = NLA_U32 },
722 [CTA_TIMEOUT_DCCP_CLOSING] = { .type = NLA_U32 },
723 [CTA_TIMEOUT_DCCP_TIMEWAIT] = { .type = NLA_U32 },
724 };
725 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
726
727 #ifdef CONFIG_SYSCTL
728 /* template, data assigned later */
729 static struct ctl_table dccp_sysctl_table[] = {
730 {
731 .procname = "nf_conntrack_dccp_timeout_request",
732 .maxlen = sizeof(unsigned int),
733 .mode = 0644,
734 .proc_handler = proc_dointvec_jiffies,
735 },
736 {
737 .procname = "nf_conntrack_dccp_timeout_respond",
738 .maxlen = sizeof(unsigned int),
739 .mode = 0644,
740 .proc_handler = proc_dointvec_jiffies,
741 },
742 {
743 .procname = "nf_conntrack_dccp_timeout_partopen",
744 .maxlen = sizeof(unsigned int),
745 .mode = 0644,
746 .proc_handler = proc_dointvec_jiffies,
747 },
748 {
749 .procname = "nf_conntrack_dccp_timeout_open",
750 .maxlen = sizeof(unsigned int),
751 .mode = 0644,
752 .proc_handler = proc_dointvec_jiffies,
753 },
754 {
755 .procname = "nf_conntrack_dccp_timeout_closereq",
756 .maxlen = sizeof(unsigned int),
757 .mode = 0644,
758 .proc_handler = proc_dointvec_jiffies,
759 },
760 {
761 .procname = "nf_conntrack_dccp_timeout_closing",
762 .maxlen = sizeof(unsigned int),
763 .mode = 0644,
764 .proc_handler = proc_dointvec_jiffies,
765 },
766 {
767 .procname = "nf_conntrack_dccp_timeout_timewait",
768 .maxlen = sizeof(unsigned int),
769 .mode = 0644,
770 .proc_handler = proc_dointvec_jiffies,
771 },
772 {
773 .procname = "nf_conntrack_dccp_loose",
774 .maxlen = sizeof(int),
775 .mode = 0644,
776 .proc_handler = proc_dointvec,
777 },
778 { }
779 };
780 #endif /* CONFIG_SYSCTL */
781
782 static int dccp_kmemdup_sysctl_table(struct net *net, struct nf_proto_net *pn,
783 struct nf_dccp_net *dn)
784 {
785 #ifdef CONFIG_SYSCTL
786 if (pn->ctl_table)
787 return 0;
788
789 pn->ctl_table = kmemdup(dccp_sysctl_table,
790 sizeof(dccp_sysctl_table),
791 GFP_KERNEL);
792 if (!pn->ctl_table)
793 return -ENOMEM;
794
795 pn->ctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST];
796 pn->ctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND];
797 pn->ctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN];
798 pn->ctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN];
799 pn->ctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ];
800 pn->ctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING];
801 pn->ctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT];
802 pn->ctl_table[7].data = &dn->dccp_loose;
803
804 /* Don't export sysctls to unprivileged users */
805 if (net->user_ns != &init_user_ns)
806 pn->ctl_table[0].procname = NULL;
807 #endif
808 return 0;
809 }
810
811 static int dccp_init_net(struct net *net)
812 {
813 struct nf_dccp_net *dn = nf_dccp_pernet(net);
814 struct nf_proto_net *pn = &dn->pn;
815
816 if (!pn->users) {
817 /* default values */
818 dn->dccp_loose = 1;
819 dn->dccp_timeout[CT_DCCP_REQUEST] = 2 * DCCP_MSL;
820 dn->dccp_timeout[CT_DCCP_RESPOND] = 4 * DCCP_MSL;
821 dn->dccp_timeout[CT_DCCP_PARTOPEN] = 4 * DCCP_MSL;
822 dn->dccp_timeout[CT_DCCP_OPEN] = 12 * 3600 * HZ;
823 dn->dccp_timeout[CT_DCCP_CLOSEREQ] = 64 * HZ;
824 dn->dccp_timeout[CT_DCCP_CLOSING] = 64 * HZ;
825 dn->dccp_timeout[CT_DCCP_TIMEWAIT] = 2 * DCCP_MSL;
826
827 /* timeouts[0] is unused, make it same as SYN_SENT so
828 * ->timeouts[0] contains 'new' timeout, like udp or icmp.
829 */
830 dn->dccp_timeout[CT_DCCP_NONE] = dn->dccp_timeout[CT_DCCP_REQUEST];
831 }
832
833 return dccp_kmemdup_sysctl_table(net, pn, dn);
834 }
835
836 static struct nf_proto_net *dccp_get_net_proto(struct net *net)
837 {
838 return &net->ct.nf_ct_proto.dccp.pn;
839 }
840
841 const struct nf_conntrack_l4proto nf_conntrack_l4proto_dccp = {
842 .l4proto = IPPROTO_DCCP,
843 .can_early_drop = dccp_can_early_drop,
844 #ifdef CONFIG_NF_CONNTRACK_PROCFS
845 .print_conntrack = dccp_print_conntrack,
846 #endif
847 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
848 .nlattr_size = DCCP_NLATTR_SIZE,
849 .to_nlattr = dccp_to_nlattr,
850 .from_nlattr = nlattr_to_dccp,
851 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
852 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size,
853 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
854 .nla_policy = nf_ct_port_nla_policy,
855 #endif
856 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
857 .ctnl_timeout = {
858 .nlattr_to_obj = dccp_timeout_nlattr_to_obj,
859 .obj_to_nlattr = dccp_timeout_obj_to_nlattr,
860 .nlattr_max = CTA_TIMEOUT_DCCP_MAX,
861 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX,
862 .nla_policy = dccp_timeout_nla_policy,
863 },
864 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
865 .init_net = dccp_init_net,
866 .get_net_proto = dccp_get_net_proto,
867 };