1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the TCP protocol. 7 * 8 * Version: @(#)tcp.h 1.0.2 04/28/93 9 * 10 * Author: Fred N. van Kempen, <[email protected]> 11 * 12 * This program is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU General Public License 14 * as published by the Free Software Foundation; either version 15 * 2 of the License, or (at your option) any later version. 16 */ 17 #ifndef _LINUX_TCP_H 18 #define _LINUX_TCP_H 19 20 #include <linux/types.h> 21 #include <asm/byteorder.h> 22 #include <linux/socket.h> 23 24 struct tcphdr { 25 __be16 source; 26 __be16 dest; 27 __be32 seq; 28 __be32 ack_seq; 29 #if defined(__LITTLE_ENDIAN_BITFIELD) 30 __u16 res1:4, 31 doff:4, 32 fin:1, 33 syn:1, 34 rst:1, 35 psh:1, 36 ack:1, 37 urg:1, 38 ece:1, 39 cwr:1; 40 #elif defined(__BIG_ENDIAN_BITFIELD) 41 __u16 doff:4, 42 res1:4, 43 cwr:1, 44 ece:1, 45 urg:1, 46 ack:1, 47 psh:1, 48 rst:1, 49 syn:1, 50 fin:1; 51 #else 52 #error "Adjust your <asm/byteorder.h> defines" 53 #endif 54 __be16 window; 55 __sum16 check; 56 __be16 urg_ptr; 57 }; 58 59 /* 60 * The union cast uses a gcc extension to avoid aliasing problems 61 * (union is compatible to any of its members) 62 * This means this part of the code is -fstrict-aliasing safe now. 63 */ 64 union tcp_word_hdr { 65 struct tcphdr hdr; 66 __be32 words[5]; 67 }; 68 69 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3]) 70 71 enum { 72 TCP_FLAG_CWR = __constant_cpu_to_be32(0x00800000), 73 TCP_FLAG_ECE = __constant_cpu_to_be32(0x00400000), 74 TCP_FLAG_URG = __constant_cpu_to_be32(0x00200000), 75 TCP_FLAG_ACK = __constant_cpu_to_be32(0x00100000), 76 TCP_FLAG_PSH = __constant_cpu_to_be32(0x00080000), 77 TCP_FLAG_RST = __constant_cpu_to_be32(0x00040000), 78 TCP_FLAG_SYN = __constant_cpu_to_be32(0x00020000), 79 TCP_FLAG_FIN = __constant_cpu_to_be32(0x00010000), 80 TCP_RESERVED_BITS = __constant_cpu_to_be32(0x0F000000), 81 TCP_DATA_OFFSET = __constant_cpu_to_be32(0xF0000000) 82 }; 83 84 /* 85 * TCP general constants 86 */ 87 #define TCP_MSS_DEFAULT 536U /* IPv4 (RFC1122, RFC2581) */ 88 #define TCP_MSS_DESIRED 1220U /* IPv6 (tunneled), EDNS0 (RFC3226) */ 89 90 /* TCP socket options */ 91 #define TCP_NODELAY 1 /* Turn off Nagle's algorithm. */ 92 #define TCP_MAXSEG 2 /* Limit MSS */ 93 #define TCP_CORK 3 /* Never send partially complete segments */ 94 #define TCP_KEEPIDLE 4 /* Start keeplives after this period */ 95 #define TCP_KEEPINTVL 5 /* Interval between keepalives */ 96 #define TCP_KEEPCNT 6 /* Number of keepalives before death */ 97 #define TCP_SYNCNT 7 /* Number of SYN retransmits */ 98 #define TCP_LINGER2 8 /* Life time of orphaned FIN-WAIT-2 state */ 99 #define TCP_DEFER_ACCEPT 9 /* Wake up listener only when data arrive */ 100 #define TCP_WINDOW_CLAMP 10 /* Bound advertised window */ 101 #define TCP_INFO 11 /* Information about this connection. */ 102 #define TCP_QUICKACK 12 /* Block/reenable quick acks */ 103 #define TCP_CONGESTION 13 /* Congestion control algorithm */ 104 #define TCP_MD5SIG 14 /* TCP MD5 Signature (RFC2385) */ 105 #define TCP_COOKIE_TRANSACTIONS 15 /* TCP Cookie Transactions */ 106 #define TCP_THIN_LINEAR_TIMEOUTS 16 /* Use linear timeouts for thin streams*/ 107 #define TCP_THIN_DUPACK 17 /* Fast retrans. after 1 dupack */ 108 #define TCP_USER_TIMEOUT 18 /* How long for loss retry before timeout */ 109 #define TCP_REPAIR 19 /* TCP sock is under repair right now */ 110 #define TCP_REPAIR_QUEUE 20 111 #define TCP_QUEUE_SEQ 21 112 #define TCP_REPAIR_OPTIONS 22 113 114 struct tcp_repair_opt { 115 __u32 opt_code; 116 __u32 opt_val; 117 }; 118 119 enum { 120 TCP_NO_QUEUE, 121 TCP_RECV_QUEUE, 122 TCP_SEND_QUEUE, 123 TCP_QUEUES_NR, 124 }; 125 126 /* for TCP_INFO socket option */ 127 #define TCPI_OPT_TIMESTAMPS 1 128 #define TCPI_OPT_SACK 2 129 #define TCPI_OPT_WSCALE 4 130 #define TCPI_OPT_ECN 8 /* ECN was negociated at TCP session init */ 131 #define TCPI_OPT_ECN_SEEN 16 /* we received at least one packet with ECT */ 132 133 enum tcp_ca_state { 134 TCP_CA_Open = 0, 135 #define TCPF_CA_Open (1<<TCP_CA_Open) 136 TCP_CA_Disorder = 1, 137 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder) 138 TCP_CA_CWR = 2, 139 #define TCPF_CA_CWR (1<<TCP_CA_CWR) 140 TCP_CA_Recovery = 3, 141 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery) 142 TCP_CA_Loss = 4 143 #define TCPF_CA_Loss (1<<TCP_CA_Loss) 144 }; 145 146 struct tcp_info { 147 __u8 tcpi_state; 148 __u8 tcpi_ca_state; 149 __u8 tcpi_retransmits; 150 __u8 tcpi_probes; 151 __u8 tcpi_backoff; 152 __u8 tcpi_options; 153 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4; 154 155 __u32 tcpi_rto; 156 __u32 tcpi_ato; 157 __u32 tcpi_snd_mss; 158 __u32 tcpi_rcv_mss; 159 160 __u32 tcpi_unacked; 161 __u32 tcpi_sacked; 162 __u32 tcpi_lost; 163 __u32 tcpi_retrans; 164 __u32 tcpi_fackets; 165 166 /* Times. */ 167 __u32 tcpi_last_data_sent; 168 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */ 169 __u32 tcpi_last_data_recv; 170 __u32 tcpi_last_ack_recv; 171 172 /* Metrics. */ 173 __u32 tcpi_pmtu; 174 __u32 tcpi_rcv_ssthresh; 175 __u32 tcpi_rtt; 176 __u32 tcpi_rttvar; 177 __u32 tcpi_snd_ssthresh; 178 __u32 tcpi_snd_cwnd; 179 __u32 tcpi_advmss; 180 __u32 tcpi_reordering; 181 182 __u32 tcpi_rcv_rtt; 183 __u32 tcpi_rcv_space; 184 185 __u32 tcpi_total_retrans; 186 }; 187 188 /* for TCP_MD5SIG socket option */ 189 #define TCP_MD5SIG_MAXKEYLEN 80 190 191 struct tcp_md5sig { 192 struct __kernel_sockaddr_storage tcpm_addr; /* address associated */ 193 __u16 __tcpm_pad1; /* zero */ 194 __u16 tcpm_keylen; /* key length */ 195 __u32 __tcpm_pad2; /* zero */ 196 __u8 tcpm_key[TCP_MD5SIG_MAXKEYLEN]; /* key (binary) */ 197 }; 198 199 /* for TCP_COOKIE_TRANSACTIONS (TCPCT) socket option */ 200 #define TCP_COOKIE_MIN 8 /* 64-bits */ 201 #define TCP_COOKIE_MAX 16 /* 128-bits */ 202 #define TCP_COOKIE_PAIR_SIZE (2*TCP_COOKIE_MAX) 203 204 /* Flags for both getsockopt and setsockopt */ 205 #define TCP_COOKIE_IN_ALWAYS (1 << 0) /* Discard SYN without cookie */ 206 #define TCP_COOKIE_OUT_NEVER (1 << 1) /* Prohibit outgoing cookies, 207 * supercedes everything. */ 208 209 /* Flags for getsockopt */ 210 #define TCP_S_DATA_IN (1 << 2) /* Was data received? */ 211 #define TCP_S_DATA_OUT (1 << 3) /* Was data sent? */ 212 213 /* TCP_COOKIE_TRANSACTIONS data */ 214 struct tcp_cookie_transactions { 215 __u16 tcpct_flags; /* see above */ 216 __u8 __tcpct_pad1; /* zero */ 217 __u8 tcpct_cookie_desired; /* bytes */ 218 __u16 tcpct_s_data_desired; /* bytes of variable data */ 219 __u16 tcpct_used; /* bytes in value */ 220 __u8 tcpct_value[TCP_MSS_DEFAULT]; 221 }; 222 223 #ifdef __KERNEL__ 224 225 #include <linux/skbuff.h> 226 #include <linux/dmaengine.h> 227 #include <net/sock.h> 228 #include <net/inet_connection_sock.h> 229 #include <net/inet_timewait_sock.h> 230 231 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb) 232 { 233 return (struct tcphdr *)skb_transport_header(skb); 234 } 235 236 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb) 237 { 238 return tcp_hdr(skb)->doff * 4; 239 } 240 241 static inline unsigned int tcp_optlen(const struct sk_buff *skb) 242 { 243 return (tcp_hdr(skb)->doff - 5) * 4; 244 } 245 246 /* TCP Fast Open */ 247 #define TCP_FASTOPEN_COOKIE_MIN 4 /* Min Fast Open Cookie size in bytes */ 248 #define TCP_FASTOPEN_COOKIE_MAX 16 /* Max Fast Open Cookie size in bytes */ 249 250 /* TCP Fast Open Cookie as stored in memory */ 251 struct tcp_fastopen_cookie { 252 s8 len; 253 u8 val[TCP_FASTOPEN_COOKIE_MAX]; 254 }; 255 256 /* This defines a selective acknowledgement block. */ 257 struct tcp_sack_block_wire { 258 __be32 start_seq; 259 __be32 end_seq; 260 }; 261 262 struct tcp_sack_block { 263 u32 start_seq; 264 u32 end_seq; 265 }; 266 267 /*These are used to set the sack_ok field in struct tcp_options_received */ 268 #define TCP_SACK_SEEN (1 << 0) /*1 = peer is SACK capable, */ 269 #define TCP_FACK_ENABLED (1 << 1) /*1 = FACK is enabled locally*/ 270 #define TCP_DSACK_SEEN (1 << 2) /*1 = DSACK was received from peer*/ 271 272 struct tcp_options_received { 273 /* PAWS/RTTM data */ 274 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */ 275 u32 ts_recent; /* Time stamp to echo next */ 276 u32 rcv_tsval; /* Time stamp value */ 277 u32 rcv_tsecr; /* Time stamp echo reply */ 278 u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */ 279 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */ 280 dsack : 1, /* D-SACK is scheduled */ 281 wscale_ok : 1, /* Wscale seen on SYN packet */ 282 sack_ok : 4, /* SACK seen on SYN packet */ 283 snd_wscale : 4, /* Window scaling received from sender */ 284 rcv_wscale : 4; /* Window scaling to send to receiver */ 285 u8 cookie_plus:6, /* bytes in authenticator/cookie option */ 286 cookie_out_never:1, 287 cookie_in_always:1; 288 u8 num_sacks; /* Number of SACK blocks */ 289 u16 user_mss; /* mss requested by user in ioctl */ 290 u16 mss_clamp; /* Maximal mss, negotiated at connection setup */ 291 }; 292 293 static inline void tcp_clear_options(struct tcp_options_received *rx_opt) 294 { 295 rx_opt->tstamp_ok = rx_opt->sack_ok = 0; 296 rx_opt->wscale_ok = rx_opt->snd_wscale = 0; 297 rx_opt->cookie_plus = 0; 298 } 299 300 /* This is the max number of SACKS that we'll generate and process. It's safe 301 * to increase this, although since: 302 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) 303 * only four options will fit in a standard TCP header */ 304 #define TCP_NUM_SACKS 4 305 306 struct tcp_cookie_values; 307 struct tcp_request_sock_ops; 308 309 struct tcp_request_sock { 310 struct inet_request_sock req; 311 #ifdef CONFIG_TCP_MD5SIG 312 /* Only used by TCP MD5 Signature so far. */ 313 const struct tcp_request_sock_ops *af_specific; 314 #endif 315 u32 rcv_isn; 316 u32 snt_isn; 317 u32 snt_synack; /* synack sent time */ 318 }; 319 320 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req) 321 { 322 return (struct tcp_request_sock *)req; 323 } 324 325 struct tcp_sock { 326 /* inet_connection_sock has to be the first member of tcp_sock */ 327 struct inet_connection_sock inet_conn; 328 u16 tcp_header_len; /* Bytes of tcp header to send */ 329 u16 xmit_size_goal_segs; /* Goal for segmenting output packets */ 330 331 /* 332 * Header prediction flags 333 * 0x5?10 << 16 + snd_wnd in net byte order 334 */ 335 __be32 pred_flags; 336 337 /* 338 * RFC793 variables by their proper names. This means you can 339 * read the code and the spec side by side (and laugh ...) 340 * See RFC793 and RFC1122. The RFC writes these in capitals. 341 */ 342 u32 rcv_nxt; /* What we want to receive next */ 343 u32 copied_seq; /* Head of yet unread data */ 344 u32 rcv_wup; /* rcv_nxt on last window update sent */ 345 u32 snd_nxt; /* Next sequence we send */ 346 347 u32 snd_una; /* First byte we want an ack for */ 348 u32 snd_sml; /* Last byte of the most recently transmitted small packet */ 349 u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ 350 u32 lsndtime; /* timestamp of last sent data packet (for restart window) */ 351 352 struct list_head tsq_node; /* anchor in tsq_tasklet.head list */ 353 unsigned long tsq_flags; 354 355 /* Data for direct copy to user */ 356 struct { 357 struct sk_buff_head prequeue; 358 struct task_struct *task; 359 struct iovec *iov; 360 int memory; 361 int len; 362 #ifdef CONFIG_NET_DMA 363 /* members for async copy */ 364 struct dma_chan *dma_chan; 365 int wakeup; 366 struct dma_pinned_list *pinned_list; 367 dma_cookie_t dma_cookie; 368 #endif 369 } ucopy; 370 371 u32 snd_wl1; /* Sequence for window update */ 372 u32 snd_wnd; /* The window we expect to receive */ 373 u32 max_window; /* Maximal window ever seen from peer */ 374 u32 mss_cache; /* Cached effective mss, not including SACKS */ 375 376 u32 window_clamp; /* Maximal window to advertise */ 377 u32 rcv_ssthresh; /* Current window clamp */ 378 379 u32 frto_highmark; /* snd_nxt when RTO occurred */ 380 u16 advmss; /* Advertised MSS */ 381 u8 frto_counter; /* Number of new acks after RTO */ 382 u8 nonagle : 4,/* Disable Nagle algorithm? */ 383 thin_lto : 1,/* Use linear timeouts for thin streams */ 384 thin_dupack : 1,/* Fast retransmit on first dupack */ 385 repair : 1, 386 unused : 1; 387 u8 repair_queue; 388 u8 do_early_retrans:1,/* Enable RFC5827 early-retransmit */ 389 early_retrans_delayed:1, /* Delayed ER timer installed */ 390 syn_data:1, /* SYN includes data */ 391 syn_fastopen:1; /* SYN includes Fast Open option */ 392 393 /* RTT measurement */ 394 u32 srtt; /* smoothed round trip time << 3 */ 395 u32 mdev; /* medium deviation */ 396 u32 mdev_max; /* maximal mdev for the last rtt period */ 397 u32 rttvar; /* smoothed mdev_max */ 398 u32 rtt_seq; /* sequence number to update rttvar */ 399 400 u32 packets_out; /* Packets which are "in flight" */ 401 u32 retrans_out; /* Retransmitted packets out */ 402 403 u16 urg_data; /* Saved octet of OOB data and control flags */ 404 u8 ecn_flags; /* ECN status bits. */ 405 u8 reordering; /* Packet reordering metric. */ 406 u32 snd_up; /* Urgent pointer */ 407 408 u8 keepalive_probes; /* num of allowed keep alive probes */ 409 /* 410 * Options received (usually on last packet, some only on SYN packets). 411 */ 412 struct tcp_options_received rx_opt; 413 414 /* 415 * Slow start and congestion control (see also Nagle, and Karn & Partridge) 416 */ 417 u32 snd_ssthresh; /* Slow start size threshold */ 418 u32 snd_cwnd; /* Sending congestion window */ 419 u32 snd_cwnd_cnt; /* Linear increase counter */ 420 u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */ 421 u32 snd_cwnd_used; 422 u32 snd_cwnd_stamp; 423 u32 prior_cwnd; /* Congestion window at start of Recovery. */ 424 u32 prr_delivered; /* Number of newly delivered packets to 425 * receiver in Recovery. */ 426 u32 prr_out; /* Total number of pkts sent during Recovery. */ 427 428 u32 rcv_wnd; /* Current receiver window */ 429 u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ 430 u32 pushed_seq; /* Last pushed seq, required to talk to windows */ 431 u32 lost_out; /* Lost packets */ 432 u32 sacked_out; /* SACK'd packets */ 433 u32 fackets_out; /* FACK'd packets */ 434 u32 tso_deferred; 435 u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */ 436 437 /* from STCP, retrans queue hinting */ 438 struct sk_buff* lost_skb_hint; 439 struct sk_buff *scoreboard_skb_hint; 440 struct sk_buff *retransmit_skb_hint; 441 442 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */ 443 444 /* SACKs data, these 2 need to be together (see tcp_options_write) */ 445 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ 446 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ 447 448 struct tcp_sack_block recv_sack_cache[4]; 449 450 struct sk_buff *highest_sack; /* skb just after the highest 451 * skb with SACKed bit set 452 * (validity guaranteed only if 453 * sacked_out > 0) 454 */ 455 456 int lost_cnt_hint; 457 u32 retransmit_high; /* L-bits may be on up to this seqno */ 458 459 u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */ 460 461 u32 prior_ssthresh; /* ssthresh saved at recovery start */ 462 u32 high_seq; /* snd_nxt at onset of congestion */ 463 464 u32 retrans_stamp; /* Timestamp of the last retransmit, 465 * also used in SYN-SENT to remember stamp of 466 * the first SYN. */ 467 u32 undo_marker; /* tracking retrans started here. */ 468 int undo_retrans; /* number of undoable retransmissions. */ 469 u32 total_retrans; /* Total retransmits for entire connection */ 470 471 u32 urg_seq; /* Seq of received urgent pointer */ 472 unsigned int keepalive_time; /* time before keep alive takes place */ 473 unsigned int keepalive_intvl; /* time interval between keep alive probes */ 474 475 int linger2; 476 477 /* Receiver side RTT estimation */ 478 struct { 479 u32 rtt; 480 u32 seq; 481 u32 time; 482 } rcv_rtt_est; 483 484 /* Receiver queue space */ 485 struct { 486 int space; 487 u32 seq; 488 u32 time; 489 } rcvq_space; 490 491 /* TCP-specific MTU probe information. */ 492 struct { 493 u32 probe_seq_start; 494 u32 probe_seq_end; 495 } mtu_probe; 496 u32 mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG 497 * while socket was owned by user. 498 */ 499 500 #ifdef CONFIG_TCP_MD5SIG 501 /* TCP AF-Specific parts; only used by MD5 Signature support so far */ 502 const struct tcp_sock_af_ops *af_specific; 503 504 /* TCP MD5 Signature Option information */ 505 struct tcp_md5sig_info __rcu *md5sig_info; 506 #endif 507 508 /* TCP fastopen related information */ 509 struct tcp_fastopen_request *fastopen_req; 510 511 /* When the cookie options are generated and exchanged, then this 512 * object holds a reference to them (cookie_values->kref). Also 513 * contains related tcp_cookie_transactions fields. 514 */ 515 struct tcp_cookie_values *cookie_values; 516 }; 517 518 enum tsq_flags { 519 TSQ_THROTTLED, 520 TSQ_QUEUED, 521 TCP_TSQ_DEFERRED, /* tcp_tasklet_func() found socket was owned */ 522 TCP_WRITE_TIMER_DEFERRED, /* tcp_write_timer() found socket was owned */ 523 TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */ 524 TCP_MTU_REDUCED_DEFERRED, /* tcp_v{4|6}_err() could not call 525 * tcp_v{4|6}_mtu_reduced() 526 */ 527 }; 528 529 static inline struct tcp_sock *tcp_sk(const struct sock *sk) 530 { 531 return (struct tcp_sock *)sk; 532 } 533 534 struct tcp_timewait_sock { 535 struct inet_timewait_sock tw_sk; 536 u32 tw_rcv_nxt; 537 u32 tw_snd_nxt; 538 u32 tw_rcv_wnd; 539 u32 tw_ts_recent; 540 long tw_ts_recent_stamp; 541 #ifdef CONFIG_TCP_MD5SIG 542 struct tcp_md5sig_key *tw_md5_key; 543 #endif 544 /* Few sockets in timewait have cookies; in that case, then this 545 * object holds a reference to them (tw_cookie_values->kref). 546 */ 547 struct tcp_cookie_values *tw_cookie_values; 548 }; 549 550 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk) 551 { 552 return (struct tcp_timewait_sock *)sk; 553 } 554 555 #endif /* __KERNEL__ */ 556 557 #endif /* _LINUX_TCP_H */ 558