1#ifndef _DCCP_H
2#define _DCCP_H
3/*
4 *  net/dccp/dccp.h
5 *
6 *  An implementation of the DCCP protocol
7 *  Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
8 *  Copyright (c) 2005-6 Ian McDonald <ian.mcdonald@jandi.co.nz>
9 *
10 *	This program is free software; you can redistribute it and/or modify it
11 *	under the terms of the GNU General Public License version 2 as
12 *	published by the Free Software Foundation.
13 */
14
15#include <linux/dccp.h>
16#include <linux/ktime.h>
17#include <net/snmp.h>
18#include <net/sock.h>
19#include <net/tcp.h>
20#include "ackvec.h"
21
22/*
23 * 	DCCP - specific warning and debugging macros.
24 */
25#define DCCP_WARN(fmt, ...)						\
26	net_warn_ratelimited("%s: " fmt, __func__, ##__VA_ARGS__)
27#define DCCP_CRIT(fmt, a...) printk(KERN_CRIT fmt " at %s:%d/%s()\n", ##a, \
28					 __FILE__, __LINE__, __func__)
29#define DCCP_BUG(a...)       do { DCCP_CRIT("BUG: " a); dump_stack(); } while(0)
30#define DCCP_BUG_ON(cond)    do { if (unlikely((cond) != 0))		   \
31				     DCCP_BUG("\"%s\" holds (exception!)", \
32					      __stringify(cond));          \
33			     } while (0)
34
35#define DCCP_PRINTK(enable, fmt, args...)	do { if (enable)	     \
36							printk(fmt, ##args); \
37						} while(0)
38#define DCCP_PR_DEBUG(enable, fmt, a...)	DCCP_PRINTK(enable, KERN_DEBUG \
39						  "%s: " fmt, __func__, ##a)
40
41#ifdef CONFIG_IP_DCCP_DEBUG
42extern bool dccp_debug;
43#define dccp_pr_debug(format, a...)	  DCCP_PR_DEBUG(dccp_debug, format, ##a)
44#define dccp_pr_debug_cat(format, a...)   DCCP_PRINTK(dccp_debug, format, ##a)
45#define dccp_debug(fmt, a...)		  dccp_pr_debug_cat(KERN_DEBUG fmt, ##a)
46#else
47#define dccp_pr_debug(format, a...)
48#define dccp_pr_debug_cat(format, a...)
49#define dccp_debug(format, a...)
50#endif
51
52extern struct inet_hashinfo dccp_hashinfo;
53
54extern struct percpu_counter dccp_orphan_count;
55
56void dccp_time_wait(struct sock *sk, int state, int timeo);
57
58/*
59 *  Set safe upper bounds for header and option length. Since Data Offset is 8
60 *  bits (RFC 4340, sec. 5.1), the total header length can never be more than
61 *  4 * 255 = 1020 bytes. The largest possible header length is 28 bytes (X=1):
62 *    - DCCP-Response with ACK Subheader and 4 bytes of Service code      OR
63 *    - DCCP-Reset    with ACK Subheader and 4 bytes of Reset Code fields
64 *  Hence a safe upper bound for the maximum option length is 1020-28 = 992
65 */
66#define MAX_DCCP_SPECIFIC_HEADER (255 * sizeof(uint32_t))
67#define DCCP_MAX_PACKET_HDR 28
68#define DCCP_MAX_OPT_LEN (MAX_DCCP_SPECIFIC_HEADER - DCCP_MAX_PACKET_HDR)
69#define MAX_DCCP_HEADER (MAX_DCCP_SPECIFIC_HEADER + MAX_HEADER)
70
71/* Upper bound for initial feature-negotiation overhead (padded to 32 bits) */
72#define DCCP_FEATNEG_OVERHEAD	 (32 * sizeof(uint32_t))
73
74#define DCCP_TIMEWAIT_LEN (60 * HZ) /* how long to wait to destroy TIME-WAIT
75				     * state, about 60 seconds */
76
77/* RFC 1122, 4.2.3.1 initial RTO value */
78#define DCCP_TIMEOUT_INIT ((unsigned int)(3 * HZ))
79
80/*
81 * The maximum back-off value for retransmissions. This is needed for
82 *  - retransmitting client-Requests (sec. 8.1.1),
83 *  - retransmitting Close/CloseReq when closing (sec. 8.3),
84 *  - feature-negotiation retransmission (sec. 6.6.3),
85 *  - Acks in client-PARTOPEN state (sec. 8.1.5).
86 */
87#define DCCP_RTO_MAX ((unsigned int)(64 * HZ))
88
89/*
90 * RTT sampling: sanity bounds and fallback RTT value from RFC 4340, section 3.4
91 */
92#define DCCP_SANE_RTT_MIN	100
93#define DCCP_FALLBACK_RTT	(USEC_PER_SEC / 5)
94#define DCCP_SANE_RTT_MAX	(3 * USEC_PER_SEC)
95
96/* sysctl variables for DCCP */
97extern int  sysctl_dccp_request_retries;
98extern int  sysctl_dccp_retries1;
99extern int  sysctl_dccp_retries2;
100extern int  sysctl_dccp_tx_qlen;
101extern int  sysctl_dccp_sync_ratelimit;
102
103/*
104 *	48-bit sequence number arithmetic (signed and unsigned)
105 */
106#define INT48_MIN	  0x800000000000LL		/* 2^47	    */
107#define UINT48_MAX	  0xFFFFFFFFFFFFLL		/* 2^48 - 1 */
108#define COMPLEMENT48(x)	 (0x1000000000000LL - (x))	/* 2^48 - x */
109#define TO_SIGNED48(x)	 (((x) < INT48_MIN)? (x) : -COMPLEMENT48( (x)))
110#define TO_UNSIGNED48(x) (((x) >= 0)?	     (x) :  COMPLEMENT48(-(x)))
111#define ADD48(a, b)	 (((a) + (b)) & UINT48_MAX)
112#define SUB48(a, b)	 ADD48((a), COMPLEMENT48(b))
113
114static inline void dccp_set_seqno(u64 *seqno, u64 value)
115{
116	*seqno = value & UINT48_MAX;
117}
118
119static inline void dccp_inc_seqno(u64 *seqno)
120{
121	*seqno = ADD48(*seqno, 1);
122}
123
124/* signed mod-2^48 distance: pos. if seqno1 < seqno2, neg. if seqno1 > seqno2 */
125static inline s64 dccp_delta_seqno(const u64 seqno1, const u64 seqno2)
126{
127	u64 delta = SUB48(seqno2, seqno1);
128
129	return TO_SIGNED48(delta);
130}
131
132/* is seq1 < seq2 ? */
133static inline int before48(const u64 seq1, const u64 seq2)
134{
135	return (s64)((seq2 << 16) - (seq1 << 16)) > 0;
136}
137
138/* is seq1 > seq2 ? */
139#define after48(seq1, seq2)	before48(seq2, seq1)
140
141/* is seq2 <= seq1 <= seq3 ? */
142static inline int between48(const u64 seq1, const u64 seq2, const u64 seq3)
143{
144	return (seq3 << 16) - (seq2 << 16) >= (seq1 << 16) - (seq2 << 16);
145}
146
147static inline u64 max48(const u64 seq1, const u64 seq2)
148{
149	return after48(seq1, seq2) ? seq1 : seq2;
150}
151
152/**
153 * dccp_loss_count - Approximate the number of lost data packets in a burst loss
154 * @s1:  last known sequence number before the loss ('hole')
155 * @s2:  first sequence number seen after the 'hole'
156 * @ndp: NDP count on packet with sequence number @s2
157 */
158static inline u64 dccp_loss_count(const u64 s1, const u64 s2, const u64 ndp)
159{
160	s64 delta = dccp_delta_seqno(s1, s2);
161
162	WARN_ON(delta < 0);
163	delta -= ndp + 1;
164
165	return delta > 0 ? delta : 0;
166}
167
168/**
169 * dccp_loss_free - Evaluate condition for data loss from RFC 4340, 7.7.1
170 */
171static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp)
172{
173	return dccp_loss_count(s1, s2, ndp) == 0;
174}
175
176enum {
177	DCCP_MIB_NUM = 0,
178	DCCP_MIB_ACTIVEOPENS,			/* ActiveOpens */
179	DCCP_MIB_ESTABRESETS,			/* EstabResets */
180	DCCP_MIB_CURRESTAB,			/* CurrEstab */
181	DCCP_MIB_OUTSEGS,			/* OutSegs */
182	DCCP_MIB_OUTRSTS,
183	DCCP_MIB_ABORTONTIMEOUT,
184	DCCP_MIB_TIMEOUTS,
185	DCCP_MIB_ABORTFAILED,
186	DCCP_MIB_PASSIVEOPENS,
187	DCCP_MIB_ATTEMPTFAILS,
188	DCCP_MIB_OUTDATAGRAMS,
189	DCCP_MIB_INERRS,
190	DCCP_MIB_OPTMANDATORYERROR,
191	DCCP_MIB_INVALIDOPT,
192	__DCCP_MIB_MAX
193};
194
195#define DCCP_MIB_MAX	__DCCP_MIB_MAX
196struct dccp_mib {
197	unsigned long	mibs[DCCP_MIB_MAX];
198};
199
200DECLARE_SNMP_STAT(struct dccp_mib, dccp_statistics);
201#define DCCP_INC_STATS(field)	    SNMP_INC_STATS(dccp_statistics, field)
202#define DCCP_INC_STATS_BH(field)    SNMP_INC_STATS_BH(dccp_statistics, field)
203#define DCCP_DEC_STATS(field)	    SNMP_DEC_STATS(dccp_statistics, field)
204
205/*
206 * 	Checksumming routines
207 */
208static inline unsigned int dccp_csum_coverage(const struct sk_buff *skb)
209{
210	const struct dccp_hdr* dh = dccp_hdr(skb);
211
212	if (dh->dccph_cscov == 0)
213		return skb->len;
214	return (dh->dccph_doff + dh->dccph_cscov - 1) * sizeof(u32);
215}
216
217static inline void dccp_csum_outgoing(struct sk_buff *skb)
218{
219	unsigned int cov = dccp_csum_coverage(skb);
220
221	if (cov >= skb->len)
222		dccp_hdr(skb)->dccph_cscov = 0;
223
224	skb->csum = skb_checksum(skb, 0, (cov > skb->len)? skb->len : cov, 0);
225}
226
227void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb);
228
229int dccp_retransmit_skb(struct sock *sk);
230
231void dccp_send_ack(struct sock *sk);
232void dccp_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
233			 struct request_sock *rsk);
234
235void dccp_send_sync(struct sock *sk, const u64 seq,
236		    const enum dccp_pkt_type pkt_type);
237
238/*
239 * TX Packet Dequeueing Interface
240 */
241void dccp_qpolicy_push(struct sock *sk, struct sk_buff *skb);
242bool dccp_qpolicy_full(struct sock *sk);
243void dccp_qpolicy_drop(struct sock *sk, struct sk_buff *skb);
244struct sk_buff *dccp_qpolicy_top(struct sock *sk);
245struct sk_buff *dccp_qpolicy_pop(struct sock *sk);
246bool dccp_qpolicy_param_ok(struct sock *sk, __be32 param);
247
248/*
249 * TX Packet Output and TX Timers
250 */
251void dccp_write_xmit(struct sock *sk);
252void dccp_write_space(struct sock *sk);
253void dccp_flush_write_queue(struct sock *sk, long *time_budget);
254
255void dccp_init_xmit_timers(struct sock *sk);
256static inline void dccp_clear_xmit_timers(struct sock *sk)
257{
258	inet_csk_clear_xmit_timers(sk);
259}
260
261unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu);
262
263const char *dccp_packet_name(const int type);
264
265void dccp_set_state(struct sock *sk, const int state);
266void dccp_done(struct sock *sk);
267
268int dccp_reqsk_init(struct request_sock *rq, struct dccp_sock const *dp,
269		    struct sk_buff const *skb);
270
271int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
272
273struct sock *dccp_create_openreq_child(struct sock *sk,
274				       const struct request_sock *req,
275				       const struct sk_buff *skb);
276
277int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
278
279struct sock *dccp_v4_request_recv_sock(struct sock *sk, struct sk_buff *skb,
280				       struct request_sock *req,
281				       struct dst_entry *dst);
282struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb,
283			    struct request_sock *req);
284
285int dccp_child_process(struct sock *parent, struct sock *child,
286		       struct sk_buff *skb);
287int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
288			   struct dccp_hdr *dh, unsigned int len);
289int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
290			 const struct dccp_hdr *dh, const unsigned int len);
291
292int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized);
293void dccp_destroy_sock(struct sock *sk);
294
295void dccp_close(struct sock *sk, long timeout);
296struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst,
297				   struct request_sock *req);
298
299int dccp_connect(struct sock *sk);
300int dccp_disconnect(struct sock *sk, int flags);
301int dccp_getsockopt(struct sock *sk, int level, int optname,
302		    char __user *optval, int __user *optlen);
303int dccp_setsockopt(struct sock *sk, int level, int optname,
304		    char __user *optval, unsigned int optlen);
305#ifdef CONFIG_COMPAT
306int compat_dccp_getsockopt(struct sock *sk, int level, int optname,
307			   char __user *optval, int __user *optlen);
308int compat_dccp_setsockopt(struct sock *sk, int level, int optname,
309			   char __user *optval, unsigned int optlen);
310#endif
311int dccp_ioctl(struct sock *sk, int cmd, unsigned long arg);
312int dccp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size);
313int dccp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
314		 int flags, int *addr_len);
315void dccp_shutdown(struct sock *sk, int how);
316int inet_dccp_listen(struct socket *sock, int backlog);
317unsigned int dccp_poll(struct file *file, struct socket *sock,
318		       poll_table *wait);
319int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
320void dccp_req_err(struct sock *sk, u64 seq);
321
322struct sk_buff *dccp_ctl_make_reset(struct sock *sk, struct sk_buff *skb);
323int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code);
324void dccp_send_close(struct sock *sk, const int active);
325int dccp_invalid_packet(struct sk_buff *skb);
326u32 dccp_sample_rtt(struct sock *sk, long delta);
327
328static inline int dccp_bad_service_code(const struct sock *sk,
329					const __be32 service)
330{
331	const struct dccp_sock *dp = dccp_sk(sk);
332
333	if (dp->dccps_service == service)
334		return 0;
335	return !dccp_list_has_service(dp->dccps_service_list, service);
336}
337
338/**
339 * dccp_skb_cb  -  DCCP per-packet control information
340 * @dccpd_type: one of %dccp_pkt_type (or unknown)
341 * @dccpd_ccval: CCVal field (5.1), see e.g. RFC 4342, 8.1
342 * @dccpd_reset_code: one of %dccp_reset_codes
343 * @dccpd_reset_data: Data1..3 fields (depend on @dccpd_reset_code)
344 * @dccpd_opt_len: total length of all options (5.8) in the packet
345 * @dccpd_seq: sequence number
346 * @dccpd_ack_seq: acknowledgment number subheader field value
347 *
348 * This is used for transmission as well as for reception.
349 */
350struct dccp_skb_cb {
351	union {
352		struct inet_skb_parm	h4;
353#if IS_ENABLED(CONFIG_IPV6)
354		struct inet6_skb_parm	h6;
355#endif
356	} header;
357	__u8  dccpd_type:4;
358	__u8  dccpd_ccval:4;
359	__u8  dccpd_reset_code,
360	      dccpd_reset_data[3];
361	__u16 dccpd_opt_len;
362	__u64 dccpd_seq;
363	__u64 dccpd_ack_seq;
364};
365
366#define DCCP_SKB_CB(__skb) ((struct dccp_skb_cb *)&((__skb)->cb[0]))
367
368/* RFC 4340, sec. 7.7 */
369static inline int dccp_non_data_packet(const struct sk_buff *skb)
370{
371	const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
372
373	return type == DCCP_PKT_ACK	 ||
374	       type == DCCP_PKT_CLOSE	 ||
375	       type == DCCP_PKT_CLOSEREQ ||
376	       type == DCCP_PKT_RESET	 ||
377	       type == DCCP_PKT_SYNC	 ||
378	       type == DCCP_PKT_SYNCACK;
379}
380
381/* RFC 4340, sec. 7.7 */
382static inline int dccp_data_packet(const struct sk_buff *skb)
383{
384	const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
385
386	return type == DCCP_PKT_DATA	 ||
387	       type == DCCP_PKT_DATAACK  ||
388	       type == DCCP_PKT_REQUEST  ||
389	       type == DCCP_PKT_RESPONSE;
390}
391
392static inline int dccp_packet_without_ack(const struct sk_buff *skb)
393{
394	const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
395
396	return type == DCCP_PKT_DATA || type == DCCP_PKT_REQUEST;
397}
398
399#define DCCP_PKT_WITHOUT_ACK_SEQ (UINT48_MAX << 2)
400
401static inline void dccp_hdr_set_seq(struct dccp_hdr *dh, const u64 gss)
402{
403	struct dccp_hdr_ext *dhx = (struct dccp_hdr_ext *)((void *)dh +
404							   sizeof(*dh));
405	dh->dccph_seq2 = 0;
406	dh->dccph_seq = htons((gss >> 32) & 0xfffff);
407	dhx->dccph_seq_low = htonl(gss & 0xffffffff);
408}
409
410static inline void dccp_hdr_set_ack(struct dccp_hdr_ack_bits *dhack,
411				    const u64 gsr)
412{
413	dhack->dccph_reserved1 = 0;
414	dhack->dccph_ack_nr_high = htons(gsr >> 32);
415	dhack->dccph_ack_nr_low  = htonl(gsr & 0xffffffff);
416}
417
418static inline void dccp_update_gsr(struct sock *sk, u64 seq)
419{
420	struct dccp_sock *dp = dccp_sk(sk);
421
422	if (after48(seq, dp->dccps_gsr))
423		dp->dccps_gsr = seq;
424	/* Sequence validity window depends on remote Sequence Window (7.5.1) */
425	dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4);
426	/*
427	 * Adjust SWL so that it is not below ISR. In contrast to RFC 4340,
428	 * 7.5.1 we perform this check beyond the initial handshake: W/W' are
429	 * always > 32, so for the first W/W' packets in the lifetime of a
430	 * connection we always have to adjust SWL.
431	 * A second reason why we are doing this is that the window depends on
432	 * the feature-remote value of Sequence Window: nothing stops the peer
433	 * from updating this value while we are busy adjusting SWL for the
434	 * first W packets (we would have to count from scratch again then).
435	 * Therefore it is safer to always make sure that the Sequence Window
436	 * is not artificially extended by a peer who grows SWL downwards by
437	 * continually updating the feature-remote Sequence-Window.
438	 * If sequence numbers wrap it is bad luck. But that will take a while
439	 * (48 bit), and this measure prevents Sequence-number attacks.
440	 */
441	if (before48(dp->dccps_swl, dp->dccps_isr))
442		dp->dccps_swl = dp->dccps_isr;
443	dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4);
444}
445
446static inline void dccp_update_gss(struct sock *sk, u64 seq)
447{
448	struct dccp_sock *dp = dccp_sk(sk);
449
450	dp->dccps_gss = seq;
451	/* Ack validity window depends on local Sequence Window value (7.5.1) */
452	dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win);
453	/* Adjust AWL so that it is not below ISS - see comment above for SWL */
454	if (before48(dp->dccps_awl, dp->dccps_iss))
455		dp->dccps_awl = dp->dccps_iss;
456	dp->dccps_awh = dp->dccps_gss;
457}
458
459static inline int dccp_ackvec_pending(const struct sock *sk)
460{
461	return dccp_sk(sk)->dccps_hc_rx_ackvec != NULL &&
462	       !dccp_ackvec_is_empty(dccp_sk(sk)->dccps_hc_rx_ackvec);
463}
464
465static inline int dccp_ack_pending(const struct sock *sk)
466{
467	return dccp_ackvec_pending(sk) || inet_csk_ack_scheduled(sk);
468}
469
470int dccp_feat_signal_nn_change(struct sock *sk, u8 feat, u64 nn_val);
471int dccp_feat_finalise_settings(struct dccp_sock *dp);
472int dccp_feat_server_ccid_dependencies(struct dccp_request_sock *dreq);
473int dccp_feat_insert_opts(struct dccp_sock*, struct dccp_request_sock*,
474			  struct sk_buff *skb);
475int dccp_feat_activate_values(struct sock *sk, struct list_head *fn);
476void dccp_feat_list_purge(struct list_head *fn_list);
477
478int dccp_insert_options(struct sock *sk, struct sk_buff *skb);
479int dccp_insert_options_rsk(struct dccp_request_sock *, struct sk_buff *);
480u32 dccp_timestamp(void);
481void dccp_timestamping_init(void);
482int dccp_insert_option(struct sk_buff *skb, unsigned char option,
483		       const void *value, unsigned char len);
484
485#ifdef CONFIG_SYSCTL
486int dccp_sysctl_init(void);
487void dccp_sysctl_exit(void);
488#else
489static inline int dccp_sysctl_init(void)
490{
491	return 0;
492}
493
494static inline void dccp_sysctl_exit(void)
495{
496}
497#endif
498
499#endif /* _DCCP_H */
500