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 extern struct percpu_counter 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
dccp_inc_seqno(u64 * seqno)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 */
dccp_delta_seqno(const u64 seqno1,const u64 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 ? */
before48(const u64 seq1,const u64 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 ? */
between48(const u64 seq1,const u64 seq2,const u64 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
max48(const u64 seq1,const u64 seq2)139 static inline u64 max48(const u64 seq1, const u64 seq2)
140 {
141 return after48(seq1, seq2) ? seq1 : seq2;
142 }
143
144 /**
145 * dccp_loss_count - Approximate the number of lost data packets in a burst loss
146 * @s1: last known sequence number before the loss ('hole')
147 * @s2: first sequence number seen after the 'hole'
148 * @ndp: NDP count on packet with sequence number @s2
149 */
dccp_loss_count(const u64 s1,const u64 s2,const u64 ndp)150 static inline u64 dccp_loss_count(const u64 s1, const u64 s2, const u64 ndp)
151 {
152 s64 delta = dccp_delta_seqno(s1, s2);
153
154 WARN_ON(delta < 0);
155 delta -= ndp + 1;
156
157 return delta > 0 ? delta : 0;
158 }
159
160 /**
161 * dccp_loss_free - Evaluate condition for data loss from RFC 4340, 7.7.1
162 */
dccp_loss_free(const u64 s1,const u64 s2,const u64 ndp)163 static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp)
164 {
165 return dccp_loss_count(s1, s2, ndp) == 0;
166 }
167
168 enum {
169 DCCP_MIB_NUM = 0,
170 DCCP_MIB_ACTIVEOPENS, /* ActiveOpens */
171 DCCP_MIB_ESTABRESETS, /* EstabResets */
172 DCCP_MIB_CURRESTAB, /* CurrEstab */
173 DCCP_MIB_OUTSEGS, /* OutSegs */
174 DCCP_MIB_OUTRSTS,
175 DCCP_MIB_ABORTONTIMEOUT,
176 DCCP_MIB_TIMEOUTS,
177 DCCP_MIB_ABORTFAILED,
178 DCCP_MIB_PASSIVEOPENS,
179 DCCP_MIB_ATTEMPTFAILS,
180 DCCP_MIB_OUTDATAGRAMS,
181 DCCP_MIB_INERRS,
182 DCCP_MIB_OPTMANDATORYERROR,
183 DCCP_MIB_INVALIDOPT,
184 __DCCP_MIB_MAX
185 };
186
187 #define DCCP_MIB_MAX __DCCP_MIB_MAX
188 struct dccp_mib {
189 unsigned long mibs[DCCP_MIB_MAX];
190 };
191
192 DECLARE_SNMP_STAT(struct dccp_mib, dccp_statistics);
193 #define DCCP_INC_STATS(field) SNMP_INC_STATS(dccp_statistics, field)
194 #define __DCCP_INC_STATS(field) __SNMP_INC_STATS(dccp_statistics, field)
195 #define DCCP_DEC_STATS(field) SNMP_DEC_STATS(dccp_statistics, field)
196
197 /*
198 * Checksumming routines
199 */
dccp_csum_coverage(const struct sk_buff * skb)200 static inline unsigned int dccp_csum_coverage(const struct sk_buff *skb)
201 {
202 const struct dccp_hdr* dh = dccp_hdr(skb);
203
204 if (dh->dccph_cscov == 0)
205 return skb->len;
206 return (dh->dccph_doff + dh->dccph_cscov - 1) * sizeof(u32);
207 }
208
dccp_csum_outgoing(struct sk_buff * skb)209 static inline void dccp_csum_outgoing(struct sk_buff *skb)
210 {
211 unsigned int cov = dccp_csum_coverage(skb);
212
213 if (cov >= skb->len)
214 dccp_hdr(skb)->dccph_cscov = 0;
215
216 skb->csum = skb_checksum(skb, 0, (cov > skb->len)? skb->len : cov, 0);
217 }
218
219 void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb);
220
221 int dccp_retransmit_skb(struct sock *sk);
222
223 void dccp_send_ack(struct sock *sk);
224 void dccp_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
225 struct request_sock *rsk);
226
227 void dccp_send_sync(struct sock *sk, const u64 seq,
228 const enum dccp_pkt_type pkt_type);
229
230 /*
231 * TX Packet Dequeueing Interface
232 */
233 void dccp_qpolicy_push(struct sock *sk, struct sk_buff *skb);
234 bool dccp_qpolicy_full(struct sock *sk);
235 void dccp_qpolicy_drop(struct sock *sk, struct sk_buff *skb);
236 struct sk_buff *dccp_qpolicy_top(struct sock *sk);
237 struct sk_buff *dccp_qpolicy_pop(struct sock *sk);
238 bool dccp_qpolicy_param_ok(struct sock *sk, __be32 param);
239
240 /*
241 * TX Packet Output and TX Timers
242 */
243 void dccp_write_xmit(struct sock *sk);
244 void dccp_write_space(struct sock *sk);
245 void dccp_flush_write_queue(struct sock *sk, long *time_budget);
246
247 void dccp_init_xmit_timers(struct sock *sk);
dccp_clear_xmit_timers(struct sock * sk)248 static inline void dccp_clear_xmit_timers(struct sock *sk)
249 {
250 inet_csk_clear_xmit_timers(sk);
251 }
252
253 unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu);
254
255 const char *dccp_packet_name(const int type);
256
257 void dccp_set_state(struct sock *sk, const int state);
258 void dccp_done(struct sock *sk);
259
260 int dccp_reqsk_init(struct request_sock *rq, struct dccp_sock const *dp,
261 struct sk_buff const *skb);
262
263 int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
264
265 struct sock *dccp_create_openreq_child(const struct sock *sk,
266 const struct request_sock *req,
267 const struct sk_buff *skb);
268
269 int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
270
271 struct sock *dccp_v4_request_recv_sock(const struct sock *sk, struct sk_buff *skb,
272 struct request_sock *req,
273 struct dst_entry *dst,
274 struct request_sock *req_unhash,
275 bool *own_req);
276 struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb,
277 struct request_sock *req);
278
279 int dccp_child_process(struct sock *parent, struct sock *child,
280 struct sk_buff *skb);
281 int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
282 struct dccp_hdr *dh, unsigned int len);
283 int dccp_rcv_established(struct sock *sk, struct sk_buff *skb,
284 const struct dccp_hdr *dh, const unsigned int len);
285
286 void dccp_destruct_common(struct sock *sk);
287 int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized);
288 void dccp_destroy_sock(struct sock *sk);
289
290 void dccp_close(struct sock *sk, long timeout);
291 struct sk_buff *dccp_make_response(const struct sock *sk, struct dst_entry *dst,
292 struct request_sock *req);
293
294 int dccp_connect(struct sock *sk);
295 int dccp_disconnect(struct sock *sk, int flags);
296 int dccp_getsockopt(struct sock *sk, int level, int optname,
297 char __user *optval, int __user *optlen);
298 int dccp_setsockopt(struct sock *sk, int level, int optname,
299 sockptr_t optval, unsigned int optlen);
300 int dccp_ioctl(struct sock *sk, int cmd, unsigned long arg);
301 int dccp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size);
302 int dccp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
303 int flags, int *addr_len);
304 void dccp_shutdown(struct sock *sk, int how);
305 int inet_dccp_listen(struct socket *sock, int backlog);
306 __poll_t dccp_poll(struct file *file, struct socket *sock,
307 poll_table *wait);
308 int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
309 void dccp_req_err(struct sock *sk, u64 seq);
310
311 struct sk_buff *dccp_ctl_make_reset(struct sock *sk, struct sk_buff *skb);
312 int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code);
313 void dccp_send_close(struct sock *sk, const int active);
314 int dccp_invalid_packet(struct sk_buff *skb);
315 u32 dccp_sample_rtt(struct sock *sk, long delta);
316
dccp_bad_service_code(const struct sock * sk,const __be32 service)317 static inline bool dccp_bad_service_code(const struct sock *sk,
318 const __be32 service)
319 {
320 const struct dccp_sock *dp = dccp_sk(sk);
321
322 if (dp->dccps_service == service)
323 return false;
324 return !dccp_list_has_service(dp->dccps_service_list, service);
325 }
326
327 /**
328 * dccp_skb_cb - DCCP per-packet control information
329 * @dccpd_type: one of %dccp_pkt_type (or unknown)
330 * @dccpd_ccval: CCVal field (5.1), see e.g. RFC 4342, 8.1
331 * @dccpd_reset_code: one of %dccp_reset_codes
332 * @dccpd_reset_data: Data1..3 fields (depend on @dccpd_reset_code)
333 * @dccpd_opt_len: total length of all options (5.8) in the packet
334 * @dccpd_seq: sequence number
335 * @dccpd_ack_seq: acknowledgment number subheader field value
336 *
337 * This is used for transmission as well as for reception.
338 */
339 struct dccp_skb_cb {
340 union {
341 struct inet_skb_parm h4;
342 #if IS_ENABLED(CONFIG_IPV6)
343 struct inet6_skb_parm h6;
344 #endif
345 } header;
346 __u8 dccpd_type:4;
347 __u8 dccpd_ccval:4;
348 __u8 dccpd_reset_code,
349 dccpd_reset_data[3];
350 __u16 dccpd_opt_len;
351 __u64 dccpd_seq;
352 __u64 dccpd_ack_seq;
353 };
354
355 #define DCCP_SKB_CB(__skb) ((struct dccp_skb_cb *)&((__skb)->cb[0]))
356
357 /* RFC 4340, sec. 7.7 */
dccp_non_data_packet(const struct sk_buff * skb)358 static inline int dccp_non_data_packet(const struct sk_buff *skb)
359 {
360 const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
361
362 return type == DCCP_PKT_ACK ||
363 type == DCCP_PKT_CLOSE ||
364 type == DCCP_PKT_CLOSEREQ ||
365 type == DCCP_PKT_RESET ||
366 type == DCCP_PKT_SYNC ||
367 type == DCCP_PKT_SYNCACK;
368 }
369
370 /* RFC 4340, sec. 7.7 */
dccp_data_packet(const struct sk_buff * skb)371 static inline int dccp_data_packet(const struct sk_buff *skb)
372 {
373 const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
374
375 return type == DCCP_PKT_DATA ||
376 type == DCCP_PKT_DATAACK ||
377 type == DCCP_PKT_REQUEST ||
378 type == DCCP_PKT_RESPONSE;
379 }
380
dccp_packet_without_ack(const struct sk_buff * skb)381 static inline int dccp_packet_without_ack(const struct sk_buff *skb)
382 {
383 const __u8 type = DCCP_SKB_CB(skb)->dccpd_type;
384
385 return type == DCCP_PKT_DATA || type == DCCP_PKT_REQUEST;
386 }
387
388 #define DCCP_PKT_WITHOUT_ACK_SEQ (UINT48_MAX << 2)
389
dccp_hdr_set_seq(struct dccp_hdr * dh,const u64 gss)390 static inline void dccp_hdr_set_seq(struct dccp_hdr *dh, const u64 gss)
391 {
392 struct dccp_hdr_ext *dhx = (struct dccp_hdr_ext *)((void *)dh +
393 sizeof(*dh));
394 dh->dccph_seq2 = 0;
395 dh->dccph_seq = htons((gss >> 32) & 0xfffff);
396 dhx->dccph_seq_low = htonl(gss & 0xffffffff);
397 }
398
dccp_hdr_set_ack(struct dccp_hdr_ack_bits * dhack,const u64 gsr)399 static inline void dccp_hdr_set_ack(struct dccp_hdr_ack_bits *dhack,
400 const u64 gsr)
401 {
402 dhack->dccph_reserved1 = 0;
403 dhack->dccph_ack_nr_high = htons(gsr >> 32);
404 dhack->dccph_ack_nr_low = htonl(gsr & 0xffffffff);
405 }
406
dccp_update_gsr(struct sock * sk,u64 seq)407 static inline void dccp_update_gsr(struct sock *sk, u64 seq)
408 {
409 struct dccp_sock *dp = dccp_sk(sk);
410
411 if (after48(seq, dp->dccps_gsr))
412 dp->dccps_gsr = seq;
413 /* Sequence validity window depends on remote Sequence Window (7.5.1) */
414 dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4);
415 /*
416 * Adjust SWL so that it is not below ISR. In contrast to RFC 4340,
417 * 7.5.1 we perform this check beyond the initial handshake: W/W' are
418 * always > 32, so for the first W/W' packets in the lifetime of a
419 * connection we always have to adjust SWL.
420 * A second reason why we are doing this is that the window depends on
421 * the feature-remote value of Sequence Window: nothing stops the peer
422 * from updating this value while we are busy adjusting SWL for the
423 * first W packets (we would have to count from scratch again then).
424 * Therefore it is safer to always make sure that the Sequence Window
425 * is not artificially extended by a peer who grows SWL downwards by
426 * continually updating the feature-remote Sequence-Window.
427 * If sequence numbers wrap it is bad luck. But that will take a while
428 * (48 bit), and this measure prevents Sequence-number attacks.
429 */
430 if (before48(dp->dccps_swl, dp->dccps_isr))
431 dp->dccps_swl = dp->dccps_isr;
432 dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4);
433 }
434
dccp_update_gss(struct sock * sk,u64 seq)435 static inline void dccp_update_gss(struct sock *sk, u64 seq)
436 {
437 struct dccp_sock *dp = dccp_sk(sk);
438
439 dp->dccps_gss = seq;
440 /* Ack validity window depends on local Sequence Window value (7.5.1) */
441 dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win);
442 /* Adjust AWL so that it is not below ISS - see comment above for SWL */
443 if (before48(dp->dccps_awl, dp->dccps_iss))
444 dp->dccps_awl = dp->dccps_iss;
445 dp->dccps_awh = dp->dccps_gss;
446 }
447
dccp_ackvec_pending(const struct sock * sk)448 static inline int dccp_ackvec_pending(const struct sock *sk)
449 {
450 return dccp_sk(sk)->dccps_hc_rx_ackvec != NULL &&
451 !dccp_ackvec_is_empty(dccp_sk(sk)->dccps_hc_rx_ackvec);
452 }
453
dccp_ack_pending(const struct sock * sk)454 static inline int dccp_ack_pending(const struct sock *sk)
455 {
456 return dccp_ackvec_pending(sk) || inet_csk_ack_scheduled(sk);
457 }
458
459 int dccp_feat_signal_nn_change(struct sock *sk, u8 feat, u64 nn_val);
460 int dccp_feat_finalise_settings(struct dccp_sock *dp);
461 int dccp_feat_server_ccid_dependencies(struct dccp_request_sock *dreq);
462 int dccp_feat_insert_opts(struct dccp_sock*, struct dccp_request_sock*,
463 struct sk_buff *skb);
464 int dccp_feat_activate_values(struct sock *sk, struct list_head *fn);
465 void dccp_feat_list_purge(struct list_head *fn_list);
466
467 int dccp_insert_options(struct sock *sk, struct sk_buff *skb);
468 int dccp_insert_options_rsk(struct dccp_request_sock *, struct sk_buff *);
469 u32 dccp_timestamp(void);
470 void dccp_timestamping_init(void);
471 int dccp_insert_option(struct sk_buff *skb, unsigned char option,
472 const void *value, unsigned char len);
473
474 #ifdef CONFIG_SYSCTL
475 int dccp_sysctl_init(void);
476 void dccp_sysctl_exit(void);
477 #else
dccp_sysctl_init(void)478 static inline int dccp_sysctl_init(void)
479 {
480 return 0;
481 }
482
dccp_sysctl_exit(void)483 static inline void dccp_sysctl_exit(void)
484 {
485 }
486 #endif
487
488 #endif /* _DCCP_H */
489