dccp.h revision 871a2c16c21b988688b4ab1a78eadd969765c0a3
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, a...) LIMIT_NETDEBUG(KERN_WARNING "%s: " fmt, \ 26 __func__, ##a) 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 int 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 56extern void 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)(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)(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/* Maximal interval between probes for local resources. */ 97#define DCCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ / 2U)) 98 99/* sysctl variables for DCCP */ 100extern int sysctl_dccp_request_retries; 101extern int sysctl_dccp_retries1; 102extern int sysctl_dccp_retries2; 103extern int sysctl_dccp_tx_qlen; 104extern int sysctl_dccp_sync_ratelimit; 105 106/* 107 * 48-bit sequence number arithmetic (signed and unsigned) 108 */ 109#define INT48_MIN 0x800000000000LL /* 2^47 */ 110#define UINT48_MAX 0xFFFFFFFFFFFFLL /* 2^48 - 1 */ 111#define COMPLEMENT48(x) (0x1000000000000LL - (x)) /* 2^48 - x */ 112#define TO_SIGNED48(x) (((x) < INT48_MIN)? (x) : -COMPLEMENT48( (x))) 113#define TO_UNSIGNED48(x) (((x) >= 0)? (x) : COMPLEMENT48(-(x))) 114#define ADD48(a, b) (((a) + (b)) & UINT48_MAX) 115#define SUB48(a, b) ADD48((a), COMPLEMENT48(b)) 116 117static inline void dccp_set_seqno(u64 *seqno, u64 value) 118{ 119 *seqno = value & UINT48_MAX; 120} 121 122static inline void dccp_inc_seqno(u64 *seqno) 123{ 124 *seqno = ADD48(*seqno, 1); 125} 126 127/* signed mod-2^48 distance: pos. if seqno1 < seqno2, neg. if seqno1 > seqno2 */ 128static inline s64 dccp_delta_seqno(const u64 seqno1, const u64 seqno2) 129{ 130 u64 delta = SUB48(seqno2, seqno1); 131 132 return TO_SIGNED48(delta); 133} 134 135/* is seq1 < seq2 ? */ 136static inline int before48(const u64 seq1, const u64 seq2) 137{ 138 return (s64)((seq2 << 16) - (seq1 << 16)) > 0; 139} 140 141/* is seq1 > seq2 ? */ 142#define after48(seq1, seq2) before48(seq2, seq1) 143 144/* is seq2 <= seq1 <= seq3 ? */ 145static inline int between48(const u64 seq1, const u64 seq2, const u64 seq3) 146{ 147 return (seq3 << 16) - (seq2 << 16) >= (seq1 << 16) - (seq2 << 16); 148} 149 150static inline u64 max48(const u64 seq1, const u64 seq2) 151{ 152 return after48(seq1, seq2) ? seq1 : seq2; 153} 154 155/** 156 * dccp_loss_count - Approximate the number of lost data packets in a burst loss 157 * @s1: last known sequence number before the loss ('hole') 158 * @s2: first sequence number seen after the 'hole' 159 * @ndp: NDP count on packet with sequence number @s2 160 */ 161static inline u64 dccp_loss_count(const u64 s1, const u64 s2, const u64 ndp) 162{ 163 s64 delta = dccp_delta_seqno(s1, s2); 164 165 WARN_ON(delta < 0); 166 delta -= ndp + 1; 167 168 return delta > 0 ? delta : 0; 169} 170 171/** 172 * dccp_loss_free - Evaluate condition for data loss from RFC 4340, 7.7.1 173 */ 174static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp) 175{ 176 return dccp_loss_count(s1, s2, ndp) == 0; 177} 178 179enum { 180 DCCP_MIB_NUM = 0, 181 DCCP_MIB_ACTIVEOPENS, /* ActiveOpens */ 182 DCCP_MIB_ESTABRESETS, /* EstabResets */ 183 DCCP_MIB_CURRESTAB, /* CurrEstab */ 184 DCCP_MIB_OUTSEGS, /* OutSegs */ 185 DCCP_MIB_OUTRSTS, 186 DCCP_MIB_ABORTONTIMEOUT, 187 DCCP_MIB_TIMEOUTS, 188 DCCP_MIB_ABORTFAILED, 189 DCCP_MIB_PASSIVEOPENS, 190 DCCP_MIB_ATTEMPTFAILS, 191 DCCP_MIB_OUTDATAGRAMS, 192 DCCP_MIB_INERRS, 193 DCCP_MIB_OPTMANDATORYERROR, 194 DCCP_MIB_INVALIDOPT, 195 __DCCP_MIB_MAX 196}; 197 198#define DCCP_MIB_MAX __DCCP_MIB_MAX 199struct dccp_mib { 200 unsigned long mibs[DCCP_MIB_MAX]; 201}; 202 203DECLARE_SNMP_STAT(struct dccp_mib, dccp_statistics); 204#define DCCP_INC_STATS(field) SNMP_INC_STATS(dccp_statistics, field) 205#define DCCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(dccp_statistics, field) 206#define DCCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(dccp_statistics, field) 207#define DCCP_DEC_STATS(field) SNMP_DEC_STATS(dccp_statistics, field) 208#define DCCP_ADD_STATS_BH(field, val) \ 209 SNMP_ADD_STATS_BH(dccp_statistics, field, val) 210#define DCCP_ADD_STATS_USER(field, val) \ 211 SNMP_ADD_STATS_USER(dccp_statistics, field, val) 212 213/* 214 * Checksumming routines 215 */ 216static inline unsigned int dccp_csum_coverage(const struct sk_buff *skb) 217{ 218 const struct dccp_hdr* dh = dccp_hdr(skb); 219 220 if (dh->dccph_cscov == 0) 221 return skb->len; 222 return (dh->dccph_doff + dh->dccph_cscov - 1) * sizeof(u32); 223} 224 225static inline void dccp_csum_outgoing(struct sk_buff *skb) 226{ 227 unsigned int cov = dccp_csum_coverage(skb); 228 229 if (cov >= skb->len) 230 dccp_hdr(skb)->dccph_cscov = 0; 231 232 skb->csum = skb_checksum(skb, 0, (cov > skb->len)? skb->len : cov, 0); 233} 234 235extern void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb); 236 237extern int dccp_retransmit_skb(struct sock *sk); 238 239extern void dccp_send_ack(struct sock *sk); 240extern void dccp_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, 241 struct request_sock *rsk); 242 243extern void dccp_send_sync(struct sock *sk, const u64 seq, 244 const enum dccp_pkt_type pkt_type); 245 246/* 247 * TX Packet Dequeueing Interface 248 */ 249extern void dccp_qpolicy_push(struct sock *sk, struct sk_buff *skb); 250extern bool dccp_qpolicy_full(struct sock *sk); 251extern void dccp_qpolicy_drop(struct sock *sk, struct sk_buff *skb); 252extern struct sk_buff *dccp_qpolicy_top(struct sock *sk); 253extern struct sk_buff *dccp_qpolicy_pop(struct sock *sk); 254 255/* 256 * TX Packet Output and TX Timers 257 */ 258extern void dccp_write_xmit(struct sock *sk); 259extern void dccp_write_space(struct sock *sk); 260extern void dccp_flush_write_queue(struct sock *sk, long *time_budget); 261 262extern void dccp_init_xmit_timers(struct sock *sk); 263static inline void dccp_clear_xmit_timers(struct sock *sk) 264{ 265 inet_csk_clear_xmit_timers(sk); 266} 267 268extern unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu); 269 270extern const char *dccp_packet_name(const int type); 271 272extern void dccp_set_state(struct sock *sk, const int state); 273extern void dccp_done(struct sock *sk); 274 275extern int dccp_reqsk_init(struct request_sock *rq, struct dccp_sock const *dp, 276 struct sk_buff const *skb); 277 278extern int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb); 279 280extern struct sock *dccp_create_openreq_child(struct sock *sk, 281 const struct request_sock *req, 282 const struct sk_buff *skb); 283 284extern int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb); 285 286extern struct sock *dccp_v4_request_recv_sock(struct sock *sk, 287 struct sk_buff *skb, 288 struct request_sock *req, 289 struct dst_entry *dst); 290extern struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb, 291 struct request_sock *req, 292 struct request_sock **prev); 293 294extern int dccp_child_process(struct sock *parent, struct sock *child, 295 struct sk_buff *skb); 296extern int dccp_rcv_state_process(struct sock *sk, struct sk_buff *skb, 297 struct dccp_hdr *dh, unsigned len); 298extern int dccp_rcv_established(struct sock *sk, struct sk_buff *skb, 299 const struct dccp_hdr *dh, const unsigned len); 300 301extern int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized); 302extern void dccp_destroy_sock(struct sock *sk); 303 304extern void dccp_close(struct sock *sk, long timeout); 305extern struct sk_buff *dccp_make_response(struct sock *sk, 306 struct dst_entry *dst, 307 struct request_sock *req); 308 309extern int dccp_connect(struct sock *sk); 310extern int dccp_disconnect(struct sock *sk, int flags); 311extern int dccp_getsockopt(struct sock *sk, int level, int optname, 312 char __user *optval, int __user *optlen); 313extern int dccp_setsockopt(struct sock *sk, int level, int optname, 314 char __user *optval, unsigned int optlen); 315#ifdef CONFIG_COMPAT 316extern int compat_dccp_getsockopt(struct sock *sk, 317 int level, int optname, 318 char __user *optval, int __user *optlen); 319extern int compat_dccp_setsockopt(struct sock *sk, 320 int level, int optname, 321 char __user *optval, unsigned int optlen); 322#endif 323extern int dccp_ioctl(struct sock *sk, int cmd, unsigned long arg); 324extern int dccp_sendmsg(struct kiocb *iocb, struct sock *sk, 325 struct msghdr *msg, size_t size); 326extern int dccp_recvmsg(struct kiocb *iocb, struct sock *sk, 327 struct msghdr *msg, size_t len, int nonblock, 328 int flags, int *addr_len); 329extern void dccp_shutdown(struct sock *sk, int how); 330extern int inet_dccp_listen(struct socket *sock, int backlog); 331extern unsigned int dccp_poll(struct file *file, struct socket *sock, 332 poll_table *wait); 333extern int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, 334 int addr_len); 335 336extern struct sk_buff *dccp_ctl_make_reset(struct sock *sk, 337 struct sk_buff *skb); 338extern int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code); 339extern void dccp_send_close(struct sock *sk, const int active); 340extern int dccp_invalid_packet(struct sk_buff *skb); 341extern u32 dccp_sample_rtt(struct sock *sk, long delta); 342 343static inline int dccp_bad_service_code(const struct sock *sk, 344 const __be32 service) 345{ 346 const struct dccp_sock *dp = dccp_sk(sk); 347 348 if (dp->dccps_service == service) 349 return 0; 350 return !dccp_list_has_service(dp->dccps_service_list, service); 351} 352 353/** 354 * dccp_skb_cb - DCCP per-packet control information 355 * @dccpd_type: one of %dccp_pkt_type (or unknown) 356 * @dccpd_ccval: CCVal field (5.1), see e.g. RFC 4342, 8.1 357 * @dccpd_reset_code: one of %dccp_reset_codes 358 * @dccpd_reset_data: Data1..3 fields (depend on @dccpd_reset_code) 359 * @dccpd_opt_len: total length of all options (5.8) in the packet 360 * @dccpd_seq: sequence number 361 * @dccpd_ack_seq: acknowledgment number subheader field value 362 * This is used for transmission as well as for reception. 363 */ 364struct dccp_skb_cb { 365 union { 366 struct inet_skb_parm h4; 367#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE) 368 struct inet6_skb_parm h6; 369#endif 370 } header; 371 __u8 dccpd_type:4; 372 __u8 dccpd_ccval:4; 373 __u8 dccpd_reset_code, 374 dccpd_reset_data[3]; 375 __u16 dccpd_opt_len; 376 __u64 dccpd_seq; 377 __u64 dccpd_ack_seq; 378}; 379 380#define DCCP_SKB_CB(__skb) ((struct dccp_skb_cb *)&((__skb)->cb[0])) 381 382/* RFC 4340, sec. 7.7 */ 383static inline int dccp_non_data_packet(const struct sk_buff *skb) 384{ 385 const __u8 type = DCCP_SKB_CB(skb)->dccpd_type; 386 387 return type == DCCP_PKT_ACK || 388 type == DCCP_PKT_CLOSE || 389 type == DCCP_PKT_CLOSEREQ || 390 type == DCCP_PKT_RESET || 391 type == DCCP_PKT_SYNC || 392 type == DCCP_PKT_SYNCACK; 393} 394 395/* RFC 4340, sec. 7.7 */ 396static inline int dccp_data_packet(const struct sk_buff *skb) 397{ 398 const __u8 type = DCCP_SKB_CB(skb)->dccpd_type; 399 400 return type == DCCP_PKT_DATA || 401 type == DCCP_PKT_DATAACK || 402 type == DCCP_PKT_REQUEST || 403 type == DCCP_PKT_RESPONSE; 404} 405 406static inline int dccp_packet_without_ack(const struct sk_buff *skb) 407{ 408 const __u8 type = DCCP_SKB_CB(skb)->dccpd_type; 409 410 return type == DCCP_PKT_DATA || type == DCCP_PKT_REQUEST; 411} 412 413#define DCCP_PKT_WITHOUT_ACK_SEQ (UINT48_MAX << 2) 414 415static inline void dccp_hdr_set_seq(struct dccp_hdr *dh, const u64 gss) 416{ 417 struct dccp_hdr_ext *dhx = (struct dccp_hdr_ext *)((void *)dh + 418 sizeof(*dh)); 419 dh->dccph_seq2 = 0; 420 dh->dccph_seq = htons((gss >> 32) & 0xfffff); 421 dhx->dccph_seq_low = htonl(gss & 0xffffffff); 422} 423 424static inline void dccp_hdr_set_ack(struct dccp_hdr_ack_bits *dhack, 425 const u64 gsr) 426{ 427 dhack->dccph_reserved1 = 0; 428 dhack->dccph_ack_nr_high = htons(gsr >> 32); 429 dhack->dccph_ack_nr_low = htonl(gsr & 0xffffffff); 430} 431 432static inline void dccp_update_gsr(struct sock *sk, u64 seq) 433{ 434 struct dccp_sock *dp = dccp_sk(sk); 435 436 dp->dccps_gsr = seq; 437 /* Sequence validity window depends on remote Sequence Window (7.5.1) */ 438 dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4); 439 /* 440 * Adjust SWL so that it is not below ISR. In contrast to RFC 4340, 441 * 7.5.1 we perform this check beyond the initial handshake: W/W' are 442 * always > 32, so for the first W/W' packets in the lifetime of a 443 * connection we always have to adjust SWL. 444 * A second reason why we are doing this is that the window depends on 445 * the feature-remote value of Sequence Window: nothing stops the peer 446 * from updating this value while we are busy adjusting SWL for the 447 * first W packets (we would have to count from scratch again then). 448 * Therefore it is safer to always make sure that the Sequence Window 449 * is not artificially extended by a peer who grows SWL downwards by 450 * continually updating the feature-remote Sequence-Window. 451 * If sequence numbers wrap it is bad luck. But that will take a while 452 * (48 bit), and this measure prevents Sequence-number attacks. 453 */ 454 if (before48(dp->dccps_swl, dp->dccps_isr)) 455 dp->dccps_swl = dp->dccps_isr; 456 dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4); 457} 458 459static inline void dccp_update_gss(struct sock *sk, u64 seq) 460{ 461 struct dccp_sock *dp = dccp_sk(sk); 462 463 dp->dccps_gss = seq; 464 /* Ack validity window depends on local Sequence Window value (7.5.1) */ 465 dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win); 466 /* Adjust AWL so that it is not below ISS - see comment above for SWL */ 467 if (before48(dp->dccps_awl, dp->dccps_iss)) 468 dp->dccps_awl = dp->dccps_iss; 469 dp->dccps_awh = dp->dccps_gss; 470} 471 472static inline int dccp_ackvec_pending(const struct sock *sk) 473{ 474 return dccp_sk(sk)->dccps_hc_rx_ackvec != NULL && 475 !dccp_ackvec_is_empty(dccp_sk(sk)->dccps_hc_rx_ackvec); 476} 477 478static inline int dccp_ack_pending(const struct sock *sk) 479{ 480 return dccp_ackvec_pending(sk) || inet_csk_ack_scheduled(sk); 481} 482 483extern int dccp_feat_finalise_settings(struct dccp_sock *dp); 484extern int dccp_feat_server_ccid_dependencies(struct dccp_request_sock *dreq); 485extern int dccp_feat_insert_opts(struct dccp_sock*, struct dccp_request_sock*, 486 struct sk_buff *skb); 487extern int dccp_feat_activate_values(struct sock *sk, struct list_head *fn); 488extern void dccp_feat_list_purge(struct list_head *fn_list); 489 490extern int dccp_insert_options(struct sock *sk, struct sk_buff *skb); 491extern int dccp_insert_options_rsk(struct dccp_request_sock*, struct sk_buff*); 492extern int dccp_insert_option_elapsed_time(struct sk_buff *skb, u32 elapsed); 493extern u32 dccp_timestamp(void); 494extern void dccp_timestamping_init(void); 495extern int dccp_insert_option(struct sk_buff *skb, unsigned char option, 496 const void *value, unsigned char len); 497 498#ifdef CONFIG_SYSCTL 499extern int dccp_sysctl_init(void); 500extern void dccp_sysctl_exit(void); 501#else 502static inline int dccp_sysctl_init(void) 503{ 504 return 0; 505} 506 507static inline void dccp_sysctl_exit(void) 508{ 509} 510#endif 511 512#endif /* _DCCP_H */ 513