xref: /linux/net/dccp/dccp.h (revision 0c7c237b1c35011ef0b8d30c1d5c20bc6ae7b69b)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 #ifndef _DCCP_H
3 #define _DCCP_H
4 /*
5  *  net/dccp/dccp.h
6  *
7  *  An implementation of the DCCP protocol
8  *  Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
9  *  Copyright (c) 2005-6 Ian McDonald <ian.mcdonald@jandi.co.nz>
10  */
11 
12 #include <linux/dccp.h>
13 #include <linux/ktime.h>
14 #include <net/snmp.h>
15 #include <net/sock.h>
16 #include <net/tcp.h>
17 #include "ackvec.h"
18 
19 /*
20  * 	DCCP - specific warning and debugging macros.
21  */
22 #define DCCP_WARN(fmt, ...)						\
23 	net_warn_ratelimited("%s: " fmt, __func__, ##__VA_ARGS__)
24 #define DCCP_CRIT(fmt, a...) printk(KERN_CRIT fmt " at %s:%d/%s()\n", ##a, \
25 					 __FILE__, __LINE__, __func__)
26 #define DCCP_BUG(a...)       do { DCCP_CRIT("BUG: " a); dump_stack(); } while(0)
27 #define DCCP_BUG_ON(cond)    do { if (unlikely((cond) != 0))		   \
28 				     DCCP_BUG("\"%s\" holds (exception!)", \
29 					      __stringify(cond));          \
30 			     } while (0)
31 
32 #define DCCP_PRINTK(enable, fmt, args...)	do { if (enable)	     \
33 							printk(fmt, ##args); \
34 						} while(0)
35 #define DCCP_PR_DEBUG(enable, fmt, a...)	DCCP_PRINTK(enable, KERN_DEBUG \
36 						  "%s: " fmt, __func__, ##a)
37 
38 #ifdef CONFIG_IP_DCCP_DEBUG
39 extern bool dccp_debug;
40 #define dccp_pr_debug(format, a...)	  DCCP_PR_DEBUG(dccp_debug, format, ##a)
41 #define dccp_pr_debug_cat(format, a...)   DCCP_PRINTK(dccp_debug, format, ##a)
42 #define dccp_debug(fmt, a...)		  dccp_pr_debug_cat(KERN_DEBUG fmt, ##a)
43 #else
44 #define dccp_pr_debug(format, a...)	  do {} while (0)
45 #define dccp_pr_debug_cat(format, a...)	  do {} while (0)
46 #define dccp_debug(format, a...)	  do {} while (0)
47 #endif
48 
49 extern struct inet_hashinfo dccp_hashinfo;
50 
51 DECLARE_PER_CPU(unsigned int, dccp_orphan_count);
52 
53 void dccp_time_wait(struct sock *sk, int state, int timeo);
54 
55 /*
56  *  Set safe upper bounds for header and option length. Since Data Offset is 8
57  *  bits (RFC 4340, sec. 5.1), the total header length can never be more than
58  *  4 * 255 = 1020 bytes. The largest possible header length is 28 bytes (X=1):
59  *    - DCCP-Response with ACK Subheader and 4 bytes of Service code      OR
60  *    - DCCP-Reset    with ACK Subheader and 4 bytes of Reset Code fields
61  *  Hence a safe upper bound for the maximum option length is 1020-28 = 992
62  */
63 #define MAX_DCCP_SPECIFIC_HEADER (255 * sizeof(uint32_t))
64 #define DCCP_MAX_PACKET_HDR 28
65 #define DCCP_MAX_OPT_LEN (MAX_DCCP_SPECIFIC_HEADER - DCCP_MAX_PACKET_HDR)
66 #define MAX_DCCP_HEADER (MAX_DCCP_SPECIFIC_HEADER + MAX_HEADER)
67 
68 /* Upper bound for initial feature-negotiation overhead (padded to 32 bits) */
69 #define DCCP_FEATNEG_OVERHEAD	 (32 * sizeof(uint32_t))
70 
71 #define DCCP_TIMEWAIT_LEN (60 * HZ) /* how long to wait to destroy TIME-WAIT
72 				     * state, about 60 seconds */
73 
74 /* RFC 1122, 4.2.3.1 initial RTO value */
75 #define DCCP_TIMEOUT_INIT ((unsigned int)(3 * HZ))
76 
77 /*
78  * The maximum back-off value for retransmissions. This is needed for
79  *  - retransmitting client-Requests (sec. 8.1.1),
80  *  - retransmitting Close/CloseReq when closing (sec. 8.3),
81  *  - feature-negotiation retransmission (sec. 6.6.3),
82  *  - Acks in client-PARTOPEN state (sec. 8.1.5).
83  */
84 #define DCCP_RTO_MAX ((unsigned int)(64 * HZ))
85 
86 /*
87  * RTT sampling: sanity bounds and fallback RTT value from RFC 4340, section 3.4
88  */
89 #define DCCP_SANE_RTT_MIN	100
90 #define DCCP_FALLBACK_RTT	(USEC_PER_SEC / 5)
91 #define DCCP_SANE_RTT_MAX	(3 * USEC_PER_SEC)
92 
93 /* sysctl variables for DCCP */
94 extern int  sysctl_dccp_request_retries;
95 extern int  sysctl_dccp_retries1;
96 extern int  sysctl_dccp_retries2;
97 extern int  sysctl_dccp_tx_qlen;
98 extern int  sysctl_dccp_sync_ratelimit;
99 
100 /*
101  *	48-bit sequence number arithmetic (signed and unsigned)
102  */
103 #define INT48_MIN	  0x800000000000LL		/* 2^47	    */
104 #define UINT48_MAX	  0xFFFFFFFFFFFFLL		/* 2^48 - 1 */
105 #define COMPLEMENT48(x)	 (0x1000000000000LL - (x))	/* 2^48 - x */
106 #define TO_SIGNED48(x)	 (((x) < INT48_MIN)? (x) : -COMPLEMENT48( (x)))
107 #define TO_UNSIGNED48(x) (((x) >= 0)?	     (x) :  COMPLEMENT48(-(x)))
108 #define ADD48(a, b)	 (((a) + (b)) & UINT48_MAX)
109 #define SUB48(a, b)	 ADD48((a), COMPLEMENT48(b))
110 
111 static inline void dccp_inc_seqno(u64 *seqno)
112 {
113 	*seqno = ADD48(*seqno, 1);
114 }
115 
116 /* signed mod-2^48 distance: pos. if seqno1 < seqno2, neg. if seqno1 > seqno2 */
117 static inline s64 dccp_delta_seqno(const u64 seqno1, const u64 seqno2)
118 {
119 	u64 delta = SUB48(seqno2, seqno1);
120 
121 	return TO_SIGNED48(delta);
122 }
123 
124 /* is seq1 < seq2 ? */
125 static inline int before48(const u64 seq1, const u64 seq2)
126 {
127 	return (s64)((seq2 << 16) - (seq1 << 16)) > 0;
128 }
129 
130 /* is seq1 > seq2 ? */
131 #define after48(seq1, seq2)	before48(seq2, seq1)
132 
133 /* is seq2 <= seq1 <= seq3 ? */
134 static inline int between48(const u64 seq1, const u64 seq2, const u64 seq3)
135 {
136 	return (seq3 << 16) - (seq2 << 16) >= (seq1 << 16) - (seq2 << 16);
137 }
138 
139 /**
140  * dccp_loss_count - Approximate the number of lost data packets in a burst loss
141  * @s1:  last known sequence number before the loss ('hole')
142  * @s2:  first sequence number seen after the 'hole'
143  * @ndp: NDP count on packet with sequence number @s2
144  */
145 static inline u64 dccp_loss_count(const u64 s1, const u64 s2, const u64 ndp)
146 {
147 	s64 delta = dccp_delta_seqno(s1, s2);
148 
149 	WARN_ON(delta < 0);
150 	delta -= ndp + 1;
151 
152 	return delta > 0 ? delta : 0;
153 }
154 
155 /**
156  * dccp_loss_free - Evaluate condition for data loss from RFC 4340, 7.7.1
157  */
158 static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp)
159 {
160 	return dccp_loss_count(s1, s2, ndp) == 0;
161 }
162 
163 enum {
164 	DCCP_MIB_NUM = 0,
165 	DCCP_MIB_ACTIVEOPENS,			/* ActiveOpens */
166 	DCCP_MIB_ESTABRESETS,			/* EstabResets */
167 	DCCP_MIB_CURRESTAB,			/* CurrEstab */
168 	DCCP_MIB_OUTSEGS,			/* OutSegs */
169 	DCCP_MIB_OUTRSTS,
170 	DCCP_MIB_ABORTONTIMEOUT,
171 	DCCP_MIB_TIMEOUTS,
172 	DCCP_MIB_ABORTFAILED,
173 	DCCP_MIB_PASSIVEOPENS,
174 	DCCP_MIB_ATTEMPTFAILS,
175 	DCCP_MIB_OUTDATAGRAMS,
176 	DCCP_MIB_INERRS,
177 	DCCP_MIB_OPTMANDATORYERROR,
178 	DCCP_MIB_INVALIDOPT,
179 	__DCCP_MIB_MAX
180 };
181 
182 #define DCCP_MIB_MAX	__DCCP_MIB_MAX
183 struct dccp_mib {
184 	unsigned long	mibs[DCCP_MIB_MAX];
185 };
186 
187 DECLARE_SNMP_STAT(struct dccp_mib, dccp_statistics);
188 #define DCCP_INC_STATS(field)	SNMP_INC_STATS(dccp_statistics, field)
189 #define __DCCP_INC_STATS(field)	__SNMP_INC_STATS(dccp_statistics, field)
190 #define DCCP_DEC_STATS(field)	SNMP_DEC_STATS(dccp_statistics, field)
191 
192 /*
193  * 	Checksumming routines
194  */
195 static inline unsigned int dccp_csum_coverage(const struct sk_buff *skb)
196 {
197 	const struct dccp_hdr* dh = dccp_hdr(skb);
198 
199 	if (dh->dccph_cscov == 0)
200 		return skb->len;
201 	return (dh->dccph_doff + dh->dccph_cscov - 1) * sizeof(u32);
202 }
203 
204 static inline void dccp_csum_outgoing(struct sk_buff *skb)
205 {
206 	unsigned int cov = dccp_csum_coverage(skb);
207 
208 	if (cov >= skb->len)
209 		dccp_hdr(skb)->dccph_cscov = 0;
210 
211 	skb->csum = skb_checksum(skb, 0, (cov > skb->len)? skb->len : cov, 0);
212 }
213 
214 void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb);
215 
216 int dccp_retransmit_skb(struct sock *sk);
217 
218 void dccp_send_ack(struct sock *sk);
219 void dccp_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
220 			 struct request_sock *rsk);
221 
222 void dccp_send_sync(struct sock *sk, const u64 seq,
223 		    const enum dccp_pkt_type pkt_type);
224 
225 /*
226  * TX Packet Dequeueing Interface
227  */
228 void dccp_qpolicy_push(struct sock *sk, struct sk_buff *skb);
229 bool dccp_qpolicy_full(struct sock *sk);
230 void dccp_qpolicy_drop(struct sock *sk, struct sk_buff *skb);
231 struct sk_buff *dccp_qpolicy_top(struct sock *sk);
232 struct sk_buff *dccp_qpolicy_pop(struct sock *sk);
233 bool dccp_qpolicy_param_ok(struct sock *sk, __be32 param);
234 
235 /*
236  * TX Packet Output and TX Timers
237  */
238 void dccp_write_xmit(struct sock *sk);
239 void dccp_write_space(struct sock *sk);
240 void dccp_flush_write_queue(struct sock *sk, long *time_budget);
241 
242 void dccp_init_xmit_timers(struct sock *sk);
243 static inline void dccp_clear_xmit_timers(struct sock *sk)
244 {
245 	inet_csk_clear_xmit_timers(sk);
246 }
247 
248 unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu);
249 
250 const char *dccp_packet_name(const int type);
251 
252 void dccp_set_state(struct sock *sk, const int state);
253 void dccp_done(struct sock *sk);
254 
255 int dccp_reqsk_init(struct request_sock *rq, struct dccp_sock const *dp,
256 		    struct sk_buff const *skb);
257 
258 int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
259 
260 struct sock *dccp_create_openreq_child(const struct sock *sk,
261 				       const struct request_sock *req,
262 				       const struct sk_buff *skb);
263 
264 int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
265 
266 struct sock *dccp_v4_request_recv_sock(const struct sock *sk, struct sk_buff *skb,
267 				       struct request_sock *req,
268 				       struct dst_entry *dst,
269 				       struct request_sock *req_unhash,
270 				       bool *own_req);
271 struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb,
272 			    struct request_sock *req);
273 
274 int dccp_child_process(struct sock *parent, struct sock *child,
275 		       struct sk_buff *skb);
276 int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
277 			   struct dccp_hdr *dh, unsigned int len);
278 int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
279 			 const struct dccp_hdr *dh, const unsigned int len);
280 
281 void dccp_destruct_common(struct sock *sk);
282 int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized);
283 void dccp_destroy_sock(struct sock *sk);
284 
285 void dccp_close(struct sock *sk, long timeout);
286 struct sk_buff *dccp_make_response(const struct sock *sk, struct dst_entry *dst,
287 				   struct request_sock *req);
288 
289 int dccp_connect(struct sock *sk);
290 int dccp_disconnect(struct sock *sk, int flags);
291 int dccp_getsockopt(struct sock *sk, int level, int optname,
292 		    char __user *optval, int __user *optlen);
293 int dccp_setsockopt(struct sock *sk, int level, int optname,
294 		    sockptr_t optval, unsigned int optlen);
295 int dccp_ioctl(struct sock *sk, int cmd, int *karg);
296 int dccp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size);
297 int dccp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags,
298 		 int *addr_len);
299 void dccp_shutdown(struct sock *sk, int how);
300 int inet_dccp_listen(struct socket *sock, int backlog);
301 __poll_t dccp_poll(struct file *file, struct socket *sock,
302 		       poll_table *wait);
303 int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
304 void dccp_req_err(struct sock *sk, u64 seq);
305 
306 struct sk_buff *dccp_ctl_make_reset(struct sock *sk, struct sk_buff *skb);
307 int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code);
308 void dccp_send_close(struct sock *sk, const int active);
309 int dccp_invalid_packet(struct sk_buff *skb);
310 u32 dccp_sample_rtt(struct sock *sk, long delta);
311 
312 static inline bool dccp_bad_service_code(const struct sock *sk,
313 					const __be32 service)
314 {
315 	const struct dccp_sock *dp = dccp_sk(sk);
316 
317 	if (dp->dccps_service == service)
318 		return false;
319 	return !dccp_list_has_service(dp->dccps_service_list, service);
320 }
321 
322 /**
323  * dccp_skb_cb  -  DCCP per-packet control information
324  * @dccpd_type: one of %dccp_pkt_type (or unknown)
325  * @dccpd_ccval: CCVal field (5.1), see e.g. RFC 4342, 8.1
326  * @dccpd_reset_code: one of %dccp_reset_codes
327  * @dccpd_reset_data: Data1..3 fields (depend on @dccpd_reset_code)
328  * @dccpd_opt_len: total length of all options (5.8) in the packet
329  * @dccpd_seq: sequence number
330  * @dccpd_ack_seq: acknowledgment number subheader field value
331  *
332  * This is used for transmission as well as for reception.
333  */
334 struct dccp_skb_cb {
335 	union {
336 		struct inet_skb_parm	h4;
337 #if IS_ENABLED(CONFIG_IPV6)
338 		struct inet6_skb_parm	h6;
339 #endif
340 	} header;
341 	__u8  dccpd_type:4;
342 	__u8  dccpd_ccval:4;
343 	__u8  dccpd_reset_code,
344 	      dccpd_reset_data[3];
345 	__u16 dccpd_opt_len;
346 	__u64 dccpd_seq;
347 	__u64 dccpd_ack_seq;
348 };
349 
350 #define DCCP_SKB_CB(__skb) ((struct dccp_skb_cb *)&((__skb)->cb[0]))
351 
352 /* RFC 4340, sec. 7.7 */
353 static inline int dccp_non_data_packet(const struct sk_buff *skb)
354 {
355 	const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
356 
357 	return type == DCCP_PKT_ACK	 ||
358 	       type == DCCP_PKT_CLOSE	 ||
359 	       type == DCCP_PKT_CLOSEREQ ||
360 	       type == DCCP_PKT_RESET	 ||
361 	       type == DCCP_PKT_SYNC	 ||
362 	       type == DCCP_PKT_SYNCACK;
363 }
364 
365 /* RFC 4340, sec. 7.7 */
366 static inline int dccp_data_packet(const struct sk_buff *skb)
367 {
368 	const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
369 
370 	return type == DCCP_PKT_DATA	 ||
371 	       type == DCCP_PKT_DATAACK  ||
372 	       type == DCCP_PKT_REQUEST  ||
373 	       type == DCCP_PKT_RESPONSE;
374 }
375 
376 static inline int dccp_packet_without_ack(const struct sk_buff *skb)
377 {
378 	const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
379 
380 	return type == DCCP_PKT_DATA || type == DCCP_PKT_REQUEST;
381 }
382 
383 #define DCCP_PKT_WITHOUT_ACK_SEQ (UINT48_MAX << 2)
384 
385 static inline void dccp_hdr_set_seq(struct dccp_hdr *dh, const u64 gss)
386 {
387 	struct dccp_hdr_ext *dhx = (struct dccp_hdr_ext *)((void *)dh +
388 							   sizeof(*dh));
389 	dh->dccph_seq2 = 0;
390 	dh->dccph_seq = htons((gss >> 32) & 0xfffff);
391 	dhx->dccph_seq_low = htonl(gss & 0xffffffff);
392 }
393 
394 static inline void dccp_hdr_set_ack(struct dccp_hdr_ack_bits *dhack,
395 				    const u64 gsr)
396 {
397 	dhack->dccph_reserved1 = 0;
398 	dhack->dccph_ack_nr_high = htons(gsr >> 32);
399 	dhack->dccph_ack_nr_low  = htonl(gsr & 0xffffffff);
400 }
401 
402 static inline void dccp_update_gsr(struct sock *sk, u64 seq)
403 {
404 	struct dccp_sock *dp = dccp_sk(sk);
405 
406 	if (after48(seq, dp->dccps_gsr))
407 		dp->dccps_gsr = seq;
408 	/* Sequence validity window depends on remote Sequence Window (7.5.1) */
409 	dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4);
410 	/*
411 	 * Adjust SWL so that it is not below ISR. In contrast to RFC 4340,
412 	 * 7.5.1 we perform this check beyond the initial handshake: W/W' are
413 	 * always > 32, so for the first W/W' packets in the lifetime of a
414 	 * connection we always have to adjust SWL.
415 	 * A second reason why we are doing this is that the window depends on
416 	 * the feature-remote value of Sequence Window: nothing stops the peer
417 	 * from updating this value while we are busy adjusting SWL for the
418 	 * first W packets (we would have to count from scratch again then).
419 	 * Therefore it is safer to always make sure that the Sequence Window
420 	 * is not artificially extended by a peer who grows SWL downwards by
421 	 * continually updating the feature-remote Sequence-Window.
422 	 * If sequence numbers wrap it is bad luck. But that will take a while
423 	 * (48 bit), and this measure prevents Sequence-number attacks.
424 	 */
425 	if (before48(dp->dccps_swl, dp->dccps_isr))
426 		dp->dccps_swl = dp->dccps_isr;
427 	dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4);
428 }
429 
430 static inline void dccp_update_gss(struct sock *sk, u64 seq)
431 {
432 	struct dccp_sock *dp = dccp_sk(sk);
433 
434 	dp->dccps_gss = seq;
435 	/* Ack validity window depends on local Sequence Window value (7.5.1) */
436 	dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win);
437 	/* Adjust AWL so that it is not below ISS - see comment above for SWL */
438 	if (before48(dp->dccps_awl, dp->dccps_iss))
439 		dp->dccps_awl = dp->dccps_iss;
440 	dp->dccps_awh = dp->dccps_gss;
441 }
442 
443 static inline int dccp_ackvec_pending(const struct sock *sk)
444 {
445 	return dccp_sk(sk)->dccps_hc_rx_ackvec != NULL &&
446 	       !dccp_ackvec_is_empty(dccp_sk(sk)->dccps_hc_rx_ackvec);
447 }
448 
449 static inline int dccp_ack_pending(const struct sock *sk)
450 {
451 	return dccp_ackvec_pending(sk) || inet_csk_ack_scheduled(sk);
452 }
453 
454 int dccp_feat_signal_nn_change(struct sock *sk, u8 feat, u64 nn_val);
455 int dccp_feat_finalise_settings(struct dccp_sock *dp);
456 int dccp_feat_server_ccid_dependencies(struct dccp_request_sock *dreq);
457 int dccp_feat_insert_opts(struct dccp_sock*, struct dccp_request_sock*,
458 			  struct sk_buff *skb);
459 int dccp_feat_activate_values(struct sock *sk, struct list_head *fn);
460 void dccp_feat_list_purge(struct list_head *fn_list);
461 
462 int dccp_insert_options(struct sock *sk, struct sk_buff *skb);
463 int dccp_insert_options_rsk(struct dccp_request_sock *, struct sk_buff *);
464 u32 dccp_timestamp(void);
465 void dccp_timestamping_init(void);
466 int dccp_insert_option(struct sk_buff *skb, unsigned char option,
467 		       const void *value, unsigned char len);
468 
469 #ifdef CONFIG_SYSCTL
470 int dccp_sysctl_init(void);
471 void dccp_sysctl_exit(void);
472 #else
473 static inline int dccp_sysctl_init(void)
474 {
475 	return 0;
476 }
477 
478 static inline void dccp_sysctl_exit(void)
479 {
480 }
481 #endif
482 
483 #endif /* _DCCP_H */
484