• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * DCCP connection tracking protocol helper
4  *
5  * Copyright (c) 2005, 2006, 2008 Patrick McHardy <kaber@trash.net>
6  */
7 #include <linux/kernel.h>
8 #include <linux/init.h>
9 #include <linux/sysctl.h>
10 #include <linux/spinlock.h>
11 #include <linux/skbuff.h>
12 #include <linux/dccp.h>
13 #include <linux/slab.h>
14 
15 #include <net/net_namespace.h>
16 #include <net/netns/generic.h>
17 
18 #include <linux/netfilter/nfnetlink_conntrack.h>
19 #include <net/netfilter/nf_conntrack.h>
20 #include <net/netfilter/nf_conntrack_l4proto.h>
21 #include <net/netfilter/nf_conntrack_ecache.h>
22 #include <net/netfilter/nf_conntrack_timeout.h>
23 #include <net/netfilter/nf_log.h>
24 
25 /* Timeouts are based on values from RFC4340:
26  *
27  * - REQUEST:
28  *
29  *   8.1.2. Client Request
30  *
31  *   A client MAY give up on its DCCP-Requests after some time
32  *   (3 minutes, for example).
33  *
34  * - RESPOND:
35  *
36  *   8.1.3. Server Response
37  *
38  *   It MAY also leave the RESPOND state for CLOSED after a timeout of
39  *   not less than 4MSL (8 minutes);
40  *
41  * - PARTOPEN:
42  *
43  *   8.1.5. Handshake Completion
44  *
45  *   If the client remains in PARTOPEN for more than 4MSL (8 minutes),
46  *   it SHOULD reset the connection with Reset Code 2, "Aborted".
47  *
48  * - OPEN:
49  *
50  *   The DCCP timestamp overflows after 11.9 hours. If the connection
51  *   stays idle this long the sequence number won't be recognized
52  *   as valid anymore.
53  *
54  * - CLOSEREQ/CLOSING:
55  *
56  *   8.3. Termination
57  *
58  *   The retransmission timer should initially be set to go off in two
59  *   round-trip times and should back off to not less than once every
60  *   64 seconds ...
61  *
62  * - TIMEWAIT:
63  *
64  *   4.3. States
65  *
66  *   A server or client socket remains in this state for 2MSL (4 minutes)
67  *   after the connection has been town down, ...
68  */
69 
70 #define DCCP_MSL (2 * 60 * HZ)
71 
72 static const char * const dccp_state_names[] = {
73 	[CT_DCCP_NONE]		= "NONE",
74 	[CT_DCCP_REQUEST]	= "REQUEST",
75 	[CT_DCCP_RESPOND]	= "RESPOND",
76 	[CT_DCCP_PARTOPEN]	= "PARTOPEN",
77 	[CT_DCCP_OPEN]		= "OPEN",
78 	[CT_DCCP_CLOSEREQ]	= "CLOSEREQ",
79 	[CT_DCCP_CLOSING]	= "CLOSING",
80 	[CT_DCCP_TIMEWAIT]	= "TIMEWAIT",
81 	[CT_DCCP_IGNORE]	= "IGNORE",
82 	[CT_DCCP_INVALID]	= "INVALID",
83 };
84 
85 #define sNO	CT_DCCP_NONE
86 #define sRQ	CT_DCCP_REQUEST
87 #define sRS	CT_DCCP_RESPOND
88 #define sPO	CT_DCCP_PARTOPEN
89 #define sOP	CT_DCCP_OPEN
90 #define sCR	CT_DCCP_CLOSEREQ
91 #define sCG	CT_DCCP_CLOSING
92 #define sTW	CT_DCCP_TIMEWAIT
93 #define sIG	CT_DCCP_IGNORE
94 #define sIV	CT_DCCP_INVALID
95 
96 /*
97  * DCCP state transition table
98  *
99  * The assumption is the same as for TCP tracking:
100  *
101  * We are the man in the middle. All the packets go through us but might
102  * get lost in transit to the destination. It is assumed that the destination
103  * can't receive segments we haven't seen.
104  *
105  * The following states exist:
106  *
107  * NONE:	Initial state, expecting Request
108  * REQUEST:	Request seen, waiting for Response from server
109  * RESPOND:	Response from server seen, waiting for Ack from client
110  * PARTOPEN:	Ack after Response seen, waiting for packet other than Response,
111  * 		Reset or Sync from server
112  * OPEN:	Packet other than Response, Reset or Sync seen
113  * CLOSEREQ:	CloseReq from server seen, expecting Close from client
114  * CLOSING:	Close seen, expecting Reset
115  * TIMEWAIT:	Reset seen
116  * IGNORE:	Not determinable whether packet is valid
117  *
118  * Some states exist only on one side of the connection: REQUEST, RESPOND,
119  * PARTOPEN, CLOSEREQ. For the other side these states are equivalent to
120  * the one it was in before.
121  *
122  * Packets are marked as ignored (sIG) if we don't know if they're valid
123  * (for example a reincarnation of a connection we didn't notice is dead
124  * already) and the server may send back a connection closing Reset or a
125  * Response. They're also used for Sync/SyncAck packets, which we don't
126  * care about.
127  */
128 static const u_int8_t
129 dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] = {
130 	[CT_DCCP_ROLE_CLIENT] = {
131 		[DCCP_PKT_REQUEST] = {
132 		/*
133 		 * sNO -> sRQ		Regular Request
134 		 * sRQ -> sRQ		Retransmitted Request or reincarnation
135 		 * sRS -> sRS		Retransmitted Request (apparently Response
136 		 * 			got lost after we saw it) or reincarnation
137 		 * sPO -> sIG		Ignore, conntrack might be out of sync
138 		 * sOP -> sIG		Ignore, conntrack might be out of sync
139 		 * sCR -> sIG		Ignore, conntrack might be out of sync
140 		 * sCG -> sIG		Ignore, conntrack might be out of sync
141 		 * sTW -> sRQ		Reincarnation
142 		 *
143 		 *	sNO, sRQ, sRS, sPO. sOP, sCR, sCG, sTW, */
144 			sRQ, sRQ, sRS, sIG, sIG, sIG, sIG, sRQ,
145 		},
146 		[DCCP_PKT_RESPONSE] = {
147 		/*
148 		 * sNO -> sIV		Invalid
149 		 * sRQ -> sIG		Ignore, might be response to ignored Request
150 		 * sRS -> sIG		Ignore, might be response to ignored Request
151 		 * sPO -> sIG		Ignore, might be response to ignored Request
152 		 * sOP -> sIG		Ignore, might be response to ignored Request
153 		 * sCR -> sIG		Ignore, might be response to ignored Request
154 		 * sCG -> sIG		Ignore, might be response to ignored Request
155 		 * sTW -> sIV		Invalid, reincarnation in reverse direction
156 		 *			goes through sRQ
157 		 *
158 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
159 			sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIV,
160 		},
161 		[DCCP_PKT_ACK] = {
162 		/*
163 		 * sNO -> sIV		No connection
164 		 * sRQ -> sIV		No connection
165 		 * sRS -> sPO		Ack for Response, move to PARTOPEN (8.1.5.)
166 		 * sPO -> sPO		Retransmitted Ack for Response, remain in PARTOPEN
167 		 * sOP -> sOP		Regular ACK, remain in OPEN
168 		 * sCR -> sCR		Ack in CLOSEREQ MAY be processed (8.3.)
169 		 * sCG -> sCG		Ack in CLOSING MAY be processed (8.3.)
170 		 * sTW -> sIV
171 		 *
172 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
173 			sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
174 		},
175 		[DCCP_PKT_DATA] = {
176 		/*
177 		 * sNO -> sIV		No connection
178 		 * sRQ -> sIV		No connection
179 		 * sRS -> sIV		No connection
180 		 * sPO -> sIV		MUST use DataAck in PARTOPEN state (8.1.5.)
181 		 * sOP -> sOP		Regular Data packet
182 		 * sCR -> sCR		Data in CLOSEREQ MAY be processed (8.3.)
183 		 * sCG -> sCG		Data in CLOSING MAY be processed (8.3.)
184 		 * sTW -> sIV
185 		 *
186 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
187 			sIV, sIV, sIV, sIV, sOP, sCR, sCG, sIV,
188 		},
189 		[DCCP_PKT_DATAACK] = {
190 		/*
191 		 * sNO -> sIV		No connection
192 		 * sRQ -> sIV		No connection
193 		 * sRS -> sPO		Ack for Response, move to PARTOPEN (8.1.5.)
194 		 * sPO -> sPO		Remain in PARTOPEN state
195 		 * sOP -> sOP		Regular DataAck packet in OPEN state
196 		 * sCR -> sCR		DataAck in CLOSEREQ MAY be processed (8.3.)
197 		 * sCG -> sCG		DataAck in CLOSING MAY be processed (8.3.)
198 		 * sTW -> sIV
199 		 *
200 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
201 			sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
202 		},
203 		[DCCP_PKT_CLOSEREQ] = {
204 		/*
205 		 * CLOSEREQ may only be sent by the server.
206 		 *
207 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
208 			sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV
209 		},
210 		[DCCP_PKT_CLOSE] = {
211 		/*
212 		 * sNO -> sIV		No connection
213 		 * sRQ -> sIV		No connection
214 		 * sRS -> sIV		No connection
215 		 * sPO -> sCG		Client-initiated close
216 		 * sOP -> sCG		Client-initiated close
217 		 * sCR -> sCG		Close in response to CloseReq (8.3.)
218 		 * sCG -> sCG		Retransmit
219 		 * sTW -> sIV		Late retransmit, already in TIME_WAIT
220 		 *
221 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
222 			sIV, sIV, sIV, sCG, sCG, sCG, sIV, sIV
223 		},
224 		[DCCP_PKT_RESET] = {
225 		/*
226 		 * sNO -> sIV		No connection
227 		 * sRQ -> sTW		Sync received or timeout, SHOULD send Reset (8.1.1.)
228 		 * sRS -> sTW		Response received without Request
229 		 * sPO -> sTW		Timeout, SHOULD send Reset (8.1.5.)
230 		 * sOP -> sTW		Connection reset
231 		 * sCR -> sTW		Connection reset
232 		 * sCG -> sTW		Connection reset
233 		 * sTW -> sIG		Ignore (don't refresh timer)
234 		 *
235 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
236 			sIV, sTW, sTW, sTW, sTW, sTW, sTW, sIG
237 		},
238 		[DCCP_PKT_SYNC] = {
239 		/*
240 		 * We currently ignore Sync packets
241 		 *
242 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
243 			sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
244 		},
245 		[DCCP_PKT_SYNCACK] = {
246 		/*
247 		 * We currently ignore SyncAck packets
248 		 *
249 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
250 			sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
251 		},
252 	},
253 	[CT_DCCP_ROLE_SERVER] = {
254 		[DCCP_PKT_REQUEST] = {
255 		/*
256 		 * sNO -> sIV		Invalid
257 		 * sRQ -> sIG		Ignore, conntrack might be out of sync
258 		 * sRS -> sIG		Ignore, conntrack might be out of sync
259 		 * sPO -> sIG		Ignore, conntrack might be out of sync
260 		 * sOP -> sIG		Ignore, conntrack might be out of sync
261 		 * sCR -> sIG		Ignore, conntrack might be out of sync
262 		 * sCG -> sIG		Ignore, conntrack might be out of sync
263 		 * sTW -> sRQ		Reincarnation, must reverse roles
264 		 *
265 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
266 			sIV, sIG, sIG, sIG, sIG, sIG, sIG, sRQ
267 		},
268 		[DCCP_PKT_RESPONSE] = {
269 		/*
270 		 * sNO -> sIV		Response without Request
271 		 * sRQ -> sRS		Response to clients Request
272 		 * sRS -> sRS		Retransmitted Response (8.1.3. SHOULD NOT)
273 		 * sPO -> sIG		Response to an ignored Request or late retransmit
274 		 * sOP -> sIG		Ignore, might be response to ignored Request
275 		 * sCR -> sIG		Ignore, might be response to ignored Request
276 		 * sCG -> sIG		Ignore, might be response to ignored Request
277 		 * sTW -> sIV		Invalid, Request from client in sTW moves to sRQ
278 		 *
279 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
280 			sIV, sRS, sRS, sIG, sIG, sIG, sIG, sIV
281 		},
282 		[DCCP_PKT_ACK] = {
283 		/*
284 		 * sNO -> sIV		No connection
285 		 * sRQ -> sIV		No connection
286 		 * sRS -> sIV		No connection
287 		 * sPO -> sOP		Enter OPEN state (8.1.5.)
288 		 * sOP -> sOP		Regular Ack in OPEN state
289 		 * sCR -> sIV		Waiting for Close from client
290 		 * sCG -> sCG		Ack in CLOSING MAY be processed (8.3.)
291 		 * sTW -> sIV
292 		 *
293 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
294 			sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
295 		},
296 		[DCCP_PKT_DATA] = {
297 		/*
298 		 * sNO -> sIV		No connection
299 		 * sRQ -> sIV		No connection
300 		 * sRS -> sIV		No connection
301 		 * sPO -> sOP		Enter OPEN state (8.1.5.)
302 		 * sOP -> sOP		Regular Data packet in OPEN state
303 		 * sCR -> sIV		Waiting for Close from client
304 		 * sCG -> sCG		Data in CLOSING MAY be processed (8.3.)
305 		 * sTW -> sIV
306 		 *
307 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
308 			sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
309 		},
310 		[DCCP_PKT_DATAACK] = {
311 		/*
312 		 * sNO -> sIV		No connection
313 		 * sRQ -> sIV		No connection
314 		 * sRS -> sIV		No connection
315 		 * sPO -> sOP		Enter OPEN state (8.1.5.)
316 		 * sOP -> sOP		Regular DataAck in OPEN state
317 		 * sCR -> sIV		Waiting for Close from client
318 		 * sCG -> sCG		Data in CLOSING MAY be processed (8.3.)
319 		 * sTW -> sIV
320 		 *
321 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
322 			sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
323 		},
324 		[DCCP_PKT_CLOSEREQ] = {
325 		/*
326 		 * sNO -> sIV		No connection
327 		 * sRQ -> sIV		No connection
328 		 * sRS -> sIV		No connection
329 		 * sPO -> sOP -> sCR	Move directly to CLOSEREQ (8.1.5.)
330 		 * sOP -> sCR		CloseReq in OPEN state
331 		 * sCR -> sCR		Retransmit
332 		 * sCG -> sCR		Simultaneous close, client sends another Close
333 		 * sTW -> sIV		Already closed
334 		 *
335 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
336 			sIV, sIV, sIV, sCR, sCR, sCR, sCR, sIV
337 		},
338 		[DCCP_PKT_CLOSE] = {
339 		/*
340 		 * sNO -> sIV		No connection
341 		 * sRQ -> sIV		No connection
342 		 * sRS -> sIV		No connection
343 		 * sPO -> sOP -> sCG	Move direcly to CLOSING
344 		 * sOP -> sCG		Move to CLOSING
345 		 * sCR -> sIV		Close after CloseReq is invalid
346 		 * sCG -> sCG		Retransmit
347 		 * sTW -> sIV		Already closed
348 		 *
349 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
350 			sIV, sIV, sIV, sCG, sCG, sIV, sCG, sIV
351 		},
352 		[DCCP_PKT_RESET] = {
353 		/*
354 		 * sNO -> sIV		No connection
355 		 * sRQ -> sTW		Reset in response to Request
356 		 * sRS -> sTW		Timeout, SHOULD send Reset (8.1.3.)
357 		 * sPO -> sTW		Timeout, SHOULD send Reset (8.1.3.)
358 		 * sOP -> sTW
359 		 * sCR -> sTW
360 		 * sCG -> sTW
361 		 * sTW -> sIG		Ignore (don't refresh timer)
362 		 *
363 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW, sTW */
364 			sIV, sTW, sTW, sTW, sTW, sTW, sTW, sTW, sIG
365 		},
366 		[DCCP_PKT_SYNC] = {
367 		/*
368 		 * We currently ignore Sync packets
369 		 *
370 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
371 			sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
372 		},
373 		[DCCP_PKT_SYNCACK] = {
374 		/*
375 		 * We currently ignore SyncAck packets
376 		 *
377 		 *	sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
378 			sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
379 		},
380 	},
381 };
382 
383 static noinline bool
dccp_new(struct nf_conn * ct,const struct sk_buff * skb,const struct dccp_hdr * dh,const struct nf_hook_state * hook_state)384 dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
385 	 const struct dccp_hdr *dh,
386 	 const struct nf_hook_state *hook_state)
387 {
388 	struct net *net = nf_ct_net(ct);
389 	struct nf_dccp_net *dn;
390 	const char *msg;
391 	u_int8_t state;
392 
393 	state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
394 	switch (state) {
395 	default:
396 		dn = nf_dccp_pernet(net);
397 		if (dn->dccp_loose == 0) {
398 			msg = "not picking up existing connection ";
399 			goto out_invalid;
400 		}
401 		break;
402 	case CT_DCCP_REQUEST:
403 		break;
404 	case CT_DCCP_INVALID:
405 		msg = "invalid state transition ";
406 		goto out_invalid;
407 	}
408 
409 	ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
410 	ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
411 	ct->proto.dccp.state = CT_DCCP_NONE;
412 	ct->proto.dccp.last_pkt = DCCP_PKT_REQUEST;
413 	ct->proto.dccp.last_dir = IP_CT_DIR_ORIGINAL;
414 	ct->proto.dccp.handshake_seq = 0;
415 	return true;
416 
417 out_invalid:
418 	nf_ct_l4proto_log_invalid(skb, ct, hook_state, "%s", msg);
419 	return false;
420 }
421 
dccp_ack_seq(const struct dccp_hdr * dh)422 static u64 dccp_ack_seq(const struct dccp_hdr *dh)
423 {
424 	const struct dccp_hdr_ack_bits *dhack;
425 
426 	dhack = (void *)dh + __dccp_basic_hdr_len(dh);
427 	return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) +
428 		     ntohl(dhack->dccph_ack_nr_low);
429 }
430 
dccp_error(const struct dccp_hdr * dh,struct sk_buff * skb,unsigned int dataoff,const struct nf_hook_state * state)431 static bool dccp_error(const struct dccp_hdr *dh,
432 		       struct sk_buff *skb, unsigned int dataoff,
433 		       const struct nf_hook_state *state)
434 {
435 	static const unsigned long require_seq48 = 1 << DCCP_PKT_REQUEST |
436 						   1 << DCCP_PKT_RESPONSE |
437 						   1 << DCCP_PKT_CLOSEREQ |
438 						   1 << DCCP_PKT_CLOSE |
439 						   1 << DCCP_PKT_RESET |
440 						   1 << DCCP_PKT_SYNC |
441 						   1 << DCCP_PKT_SYNCACK;
442 	unsigned int dccp_len = skb->len - dataoff;
443 	unsigned int cscov;
444 	const char *msg;
445 	u8 type;
446 
447 	BUILD_BUG_ON(DCCP_PKT_INVALID >= BITS_PER_LONG);
448 
449 	if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
450 	    dh->dccph_doff * 4 > dccp_len) {
451 		msg = "nf_ct_dccp: truncated/malformed packet ";
452 		goto out_invalid;
453 	}
454 
455 	cscov = dccp_len;
456 	if (dh->dccph_cscov) {
457 		cscov = (dh->dccph_cscov - 1) * 4;
458 		if (cscov > dccp_len) {
459 			msg = "nf_ct_dccp: bad checksum coverage ";
460 			goto out_invalid;
461 		}
462 	}
463 
464 	if (state->hook == NF_INET_PRE_ROUTING &&
465 	    state->net->ct.sysctl_checksum &&
466 	    nf_checksum_partial(skb, state->hook, dataoff, cscov,
467 				IPPROTO_DCCP, state->pf)) {
468 		msg = "nf_ct_dccp: bad checksum ";
469 		goto out_invalid;
470 	}
471 
472 	type = dh->dccph_type;
473 	if (type >= DCCP_PKT_INVALID) {
474 		msg = "nf_ct_dccp: reserved packet type ";
475 		goto out_invalid;
476 	}
477 
478 	if (test_bit(type, &require_seq48) && !dh->dccph_x) {
479 		msg = "nf_ct_dccp: type lacks 48bit sequence numbers";
480 		goto out_invalid;
481 	}
482 
483 	return false;
484 out_invalid:
485 	nf_l4proto_log_invalid(skb, state, IPPROTO_DCCP, "%s", msg);
486 	return true;
487 }
488 
489 struct nf_conntrack_dccp_buf {
490 	struct dccp_hdr dh;	 /* generic header part */
491 	struct dccp_hdr_ext ext; /* optional depending dh->dccph_x */
492 	union {			 /* depends on header type */
493 		struct dccp_hdr_ack_bits ack;
494 		struct dccp_hdr_request req;
495 		struct dccp_hdr_response response;
496 		struct dccp_hdr_reset rst;
497 	} u;
498 };
499 
500 static struct dccp_hdr *
dccp_header_pointer(const struct sk_buff * skb,int offset,const struct dccp_hdr * dh,struct nf_conntrack_dccp_buf * buf)501 dccp_header_pointer(const struct sk_buff *skb, int offset, const struct dccp_hdr *dh,
502 		    struct nf_conntrack_dccp_buf *buf)
503 {
504 	unsigned int hdrlen = __dccp_hdr_len(dh);
505 
506 	if (hdrlen > sizeof(*buf))
507 		return NULL;
508 
509 	return skb_header_pointer(skb, offset, hdrlen, buf);
510 }
511 
nf_conntrack_dccp_packet(struct nf_conn * ct,struct sk_buff * skb,unsigned int dataoff,enum ip_conntrack_info ctinfo,const struct nf_hook_state * state)512 int nf_conntrack_dccp_packet(struct nf_conn *ct, struct sk_buff *skb,
513 			     unsigned int dataoff,
514 			     enum ip_conntrack_info ctinfo,
515 			     const struct nf_hook_state *state)
516 {
517 	enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
518 	struct nf_conntrack_dccp_buf _dh;
519 	u_int8_t type, old_state, new_state;
520 	enum ct_dccp_roles role;
521 	unsigned int *timeouts;
522 	struct dccp_hdr *dh;
523 
524 	dh = skb_header_pointer(skb, dataoff, sizeof(*dh), &_dh.dh);
525 	if (!dh)
526 		return NF_DROP;
527 
528 	if (dccp_error(dh, skb, dataoff, state))
529 		return -NF_ACCEPT;
530 
531 	/* pull again, including possible 48 bit sequences and subtype header */
532 	dh = dccp_header_pointer(skb, dataoff, dh, &_dh);
533 	if (!dh)
534 		return NF_DROP;
535 
536 	type = dh->dccph_type;
537 	if (!nf_ct_is_confirmed(ct) && !dccp_new(ct, skb, dh, state))
538 		return -NF_ACCEPT;
539 
540 	if (type == DCCP_PKT_RESET &&
541 	    !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
542 		/* Tear down connection immediately if only reply is a RESET */
543 		nf_ct_kill_acct(ct, ctinfo, skb);
544 		return NF_ACCEPT;
545 	}
546 
547 	spin_lock_bh(&ct->lock);
548 
549 	role = ct->proto.dccp.role[dir];
550 	old_state = ct->proto.dccp.state;
551 	new_state = dccp_state_table[role][type][old_state];
552 
553 	switch (new_state) {
554 	case CT_DCCP_REQUEST:
555 		if (old_state == CT_DCCP_TIMEWAIT &&
556 		    role == CT_DCCP_ROLE_SERVER) {
557 			/* Reincarnation in the reverse direction: reopen and
558 			 * reverse client/server roles. */
559 			ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
560 			ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
561 		}
562 		break;
563 	case CT_DCCP_RESPOND:
564 		if (old_state == CT_DCCP_REQUEST)
565 			ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
566 		break;
567 	case CT_DCCP_PARTOPEN:
568 		if (old_state == CT_DCCP_RESPOND &&
569 		    type == DCCP_PKT_ACK &&
570 		    dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
571 			set_bit(IPS_ASSURED_BIT, &ct->status);
572 		break;
573 	case CT_DCCP_IGNORE:
574 		/*
575 		 * Connection tracking might be out of sync, so we ignore
576 		 * packets that might establish a new connection and resync
577 		 * if the server responds with a valid Response.
578 		 */
579 		if (ct->proto.dccp.last_dir == !dir &&
580 		    ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
581 		    type == DCCP_PKT_RESPONSE) {
582 			ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
583 			ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
584 			ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
585 			new_state = CT_DCCP_RESPOND;
586 			break;
587 		}
588 		ct->proto.dccp.last_dir = dir;
589 		ct->proto.dccp.last_pkt = type;
590 
591 		spin_unlock_bh(&ct->lock);
592 		nf_ct_l4proto_log_invalid(skb, ct, state, "%s", "invalid packet");
593 		return NF_ACCEPT;
594 	case CT_DCCP_INVALID:
595 		spin_unlock_bh(&ct->lock);
596 		nf_ct_l4proto_log_invalid(skb, ct, state, "%s", "invalid state transition");
597 		return -NF_ACCEPT;
598 	}
599 
600 	ct->proto.dccp.last_dir = dir;
601 	ct->proto.dccp.last_pkt = type;
602 	ct->proto.dccp.state = new_state;
603 	spin_unlock_bh(&ct->lock);
604 
605 	if (new_state != old_state)
606 		nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
607 
608 	timeouts = nf_ct_timeout_lookup(ct);
609 	if (!timeouts)
610 		timeouts = nf_dccp_pernet(nf_ct_net(ct))->dccp_timeout;
611 	nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]);
612 
613 	return NF_ACCEPT;
614 }
615 
dccp_can_early_drop(const struct nf_conn * ct)616 static bool dccp_can_early_drop(const struct nf_conn *ct)
617 {
618 	switch (ct->proto.dccp.state) {
619 	case CT_DCCP_CLOSEREQ:
620 	case CT_DCCP_CLOSING:
621 	case CT_DCCP_TIMEWAIT:
622 		return true;
623 	default:
624 		break;
625 	}
626 
627 	return false;
628 }
629 
630 #ifdef CONFIG_NF_CONNTRACK_PROCFS
dccp_print_conntrack(struct seq_file * s,struct nf_conn * ct)631 static void dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
632 {
633 	seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
634 }
635 #endif
636 
637 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
dccp_to_nlattr(struct sk_buff * skb,struct nlattr * nla,struct nf_conn * ct,bool destroy)638 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
639 			  struct nf_conn *ct, bool destroy)
640 {
641 	struct nlattr *nest_parms;
642 
643 	spin_lock_bh(&ct->lock);
644 	nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP);
645 	if (!nest_parms)
646 		goto nla_put_failure;
647 	if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state))
648 		goto nla_put_failure;
649 
650 	if (destroy)
651 		goto skip_state;
652 
653 	if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE,
654 		       ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) ||
655 	    nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ,
656 			 cpu_to_be64(ct->proto.dccp.handshake_seq),
657 			 CTA_PROTOINFO_DCCP_PAD))
658 		goto nla_put_failure;
659 skip_state:
660 	nla_nest_end(skb, nest_parms);
661 	spin_unlock_bh(&ct->lock);
662 
663 	return 0;
664 
665 nla_put_failure:
666 	spin_unlock_bh(&ct->lock);
667 	return -1;
668 }
669 
670 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
671 	[CTA_PROTOINFO_DCCP_STATE]	= { .type = NLA_U8 },
672 	[CTA_PROTOINFO_DCCP_ROLE]	= { .type = NLA_U8 },
673 	[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ] = { .type = NLA_U64 },
674 	[CTA_PROTOINFO_DCCP_PAD]	= { .type = NLA_UNSPEC },
675 };
676 
677 #define DCCP_NLATTR_SIZE ( \
678 	NLA_ALIGN(NLA_HDRLEN + 1) + \
679 	NLA_ALIGN(NLA_HDRLEN + 1) + \
680 	NLA_ALIGN(NLA_HDRLEN + sizeof(u64)) + \
681 	NLA_ALIGN(NLA_HDRLEN + 0))
682 
nlattr_to_dccp(struct nlattr * cda[],struct nf_conn * ct)683 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
684 {
685 	struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
686 	struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
687 	int err;
688 
689 	if (!attr)
690 		return 0;
691 
692 	err = nla_parse_nested_deprecated(tb, CTA_PROTOINFO_DCCP_MAX, attr,
693 					  dccp_nla_policy, NULL);
694 	if (err < 0)
695 		return err;
696 
697 	if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
698 	    !tb[CTA_PROTOINFO_DCCP_ROLE] ||
699 	    nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX ||
700 	    nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) {
701 		return -EINVAL;
702 	}
703 
704 	spin_lock_bh(&ct->lock);
705 	ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
706 	if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) {
707 		ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
708 		ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
709 	} else {
710 		ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_SERVER;
711 		ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_CLIENT;
712 	}
713 	if (tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]) {
714 		ct->proto.dccp.handshake_seq =
715 		be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]));
716 	}
717 	spin_unlock_bh(&ct->lock);
718 	return 0;
719 }
720 #endif
721 
722 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
723 
724 #include <linux/netfilter/nfnetlink.h>
725 #include <linux/netfilter/nfnetlink_cttimeout.h>
726 
dccp_timeout_nlattr_to_obj(struct nlattr * tb[],struct net * net,void * data)727 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[],
728 				      struct net *net, void *data)
729 {
730 	struct nf_dccp_net *dn = nf_dccp_pernet(net);
731 	unsigned int *timeouts = data;
732 	int i;
733 
734 	if (!timeouts)
735 		 timeouts = dn->dccp_timeout;
736 
737 	/* set default DCCP timeouts. */
738 	for (i=0; i<CT_DCCP_MAX; i++)
739 		timeouts[i] = dn->dccp_timeout[i];
740 
741 	/* there's a 1:1 mapping between attributes and protocol states. */
742 	for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
743 		if (tb[i]) {
744 			timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ;
745 		}
746 	}
747 
748 	timeouts[CTA_TIMEOUT_DCCP_UNSPEC] = timeouts[CTA_TIMEOUT_DCCP_REQUEST];
749 	return 0;
750 }
751 
752 static int
dccp_timeout_obj_to_nlattr(struct sk_buff * skb,const void * data)753 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
754 {
755         const unsigned int *timeouts = data;
756 	int i;
757 
758 	for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) {
759 		if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ)))
760 			goto nla_put_failure;
761 	}
762 	return 0;
763 
764 nla_put_failure:
765 	return -ENOSPC;
766 }
767 
768 static const struct nla_policy
769 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = {
770 	[CTA_TIMEOUT_DCCP_REQUEST]	= { .type = NLA_U32 },
771 	[CTA_TIMEOUT_DCCP_RESPOND]	= { .type = NLA_U32 },
772 	[CTA_TIMEOUT_DCCP_PARTOPEN]	= { .type = NLA_U32 },
773 	[CTA_TIMEOUT_DCCP_OPEN]		= { .type = NLA_U32 },
774 	[CTA_TIMEOUT_DCCP_CLOSEREQ]	= { .type = NLA_U32 },
775 	[CTA_TIMEOUT_DCCP_CLOSING]	= { .type = NLA_U32 },
776 	[CTA_TIMEOUT_DCCP_TIMEWAIT]	= { .type = NLA_U32 },
777 };
778 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
779 
nf_conntrack_dccp_init_net(struct net * net)780 void nf_conntrack_dccp_init_net(struct net *net)
781 {
782 	struct nf_dccp_net *dn = nf_dccp_pernet(net);
783 
784 	/* default values */
785 	dn->dccp_loose = 1;
786 	dn->dccp_timeout[CT_DCCP_REQUEST]	= 2 * DCCP_MSL;
787 	dn->dccp_timeout[CT_DCCP_RESPOND]	= 4 * DCCP_MSL;
788 	dn->dccp_timeout[CT_DCCP_PARTOPEN]	= 4 * DCCP_MSL;
789 	dn->dccp_timeout[CT_DCCP_OPEN]		= 12 * 3600 * HZ;
790 	dn->dccp_timeout[CT_DCCP_CLOSEREQ]	= 64 * HZ;
791 	dn->dccp_timeout[CT_DCCP_CLOSING]	= 64 * HZ;
792 	dn->dccp_timeout[CT_DCCP_TIMEWAIT]	= 2 * DCCP_MSL;
793 
794 	/* timeouts[0] is unused, make it same as SYN_SENT so
795 	 * ->timeouts[0] contains 'new' timeout, like udp or icmp.
796 	 */
797 	dn->dccp_timeout[CT_DCCP_NONE] = dn->dccp_timeout[CT_DCCP_REQUEST];
798 }
799 
800 const struct nf_conntrack_l4proto nf_conntrack_l4proto_dccp = {
801 	.l4proto		= IPPROTO_DCCP,
802 	.can_early_drop		= dccp_can_early_drop,
803 #ifdef CONFIG_NF_CONNTRACK_PROCFS
804 	.print_conntrack	= dccp_print_conntrack,
805 #endif
806 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
807 	.nlattr_size		= DCCP_NLATTR_SIZE,
808 	.to_nlattr		= dccp_to_nlattr,
809 	.from_nlattr		= nlattr_to_dccp,
810 	.tuple_to_nlattr	= nf_ct_port_tuple_to_nlattr,
811 	.nlattr_tuple_size	= nf_ct_port_nlattr_tuple_size,
812 	.nlattr_to_tuple	= nf_ct_port_nlattr_to_tuple,
813 	.nla_policy		= nf_ct_port_nla_policy,
814 #endif
815 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
816 	.ctnl_timeout		= {
817 		.nlattr_to_obj	= dccp_timeout_nlattr_to_obj,
818 		.obj_to_nlattr	= dccp_timeout_obj_to_nlattr,
819 		.nlattr_max	= CTA_TIMEOUT_DCCP_MAX,
820 		.obj_size	= sizeof(unsigned int) * CT_DCCP_MAX,
821 		.nla_policy	= dccp_timeout_nla_policy,
822 	},
823 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
824 };
825