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 = __cpu_to_be32(0x00800000), 73 TCP_FLAG_ECE = __cpu_to_be32(0x00400000), 74 TCP_FLAG_URG = __cpu_to_be32(0x00200000), 75 TCP_FLAG_ACK = __cpu_to_be32(0x00100000), 76 TCP_FLAG_PSH = __cpu_to_be32(0x00080000), 77 TCP_FLAG_RST = __cpu_to_be32(0x00040000), 78 TCP_FLAG_SYN = __cpu_to_be32(0x00020000), 79 TCP_FLAG_FIN = __cpu_to_be32(0x00010000), 80 TCP_RESERVED_BITS = __cpu_to_be32(0x0F000000), 81 TCP_DATA_OFFSET = __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 */ 106 107 /* for TCP_INFO socket option */ 108 #define TCPI_OPT_TIMESTAMPS 1 109 #define TCPI_OPT_SACK 2 110 #define TCPI_OPT_WSCALE 4 111 #define TCPI_OPT_ECN 8 112 113 enum tcp_ca_state { 114 TCP_CA_Open = 0, 115 #define TCPF_CA_Open (1<<TCP_CA_Open) 116 TCP_CA_Disorder = 1, 117 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder) 118 TCP_CA_CWR = 2, 119 #define TCPF_CA_CWR (1<<TCP_CA_CWR) 120 TCP_CA_Recovery = 3, 121 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery) 122 TCP_CA_Loss = 4 123 #define TCPF_CA_Loss (1<<TCP_CA_Loss) 124 }; 125 126 struct tcp_info { 127 __u8 tcpi_state; 128 __u8 tcpi_ca_state; 129 __u8 tcpi_retransmits; 130 __u8 tcpi_probes; 131 __u8 tcpi_backoff; 132 __u8 tcpi_options; 133 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4; 134 135 __u32 tcpi_rto; 136 __u32 tcpi_ato; 137 __u32 tcpi_snd_mss; 138 __u32 tcpi_rcv_mss; 139 140 __u32 tcpi_unacked; 141 __u32 tcpi_sacked; 142 __u32 tcpi_lost; 143 __u32 tcpi_retrans; 144 __u32 tcpi_fackets; 145 146 /* Times. */ 147 __u32 tcpi_last_data_sent; 148 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */ 149 __u32 tcpi_last_data_recv; 150 __u32 tcpi_last_ack_recv; 151 152 /* Metrics. */ 153 __u32 tcpi_pmtu; 154 __u32 tcpi_rcv_ssthresh; 155 __u32 tcpi_rtt; 156 __u32 tcpi_rttvar; 157 __u32 tcpi_snd_ssthresh; 158 __u32 tcpi_snd_cwnd; 159 __u32 tcpi_advmss; 160 __u32 tcpi_reordering; 161 162 __u32 tcpi_rcv_rtt; 163 __u32 tcpi_rcv_space; 164 165 __u32 tcpi_total_retrans; 166 }; 167 168 /* for TCP_MD5SIG socket option */ 169 #define TCP_MD5SIG_MAXKEYLEN 80 170 171 struct tcp_md5sig { 172 struct __kernel_sockaddr_storage tcpm_addr; /* address associated */ 173 __u16 __tcpm_pad1; /* zero */ 174 __u16 tcpm_keylen; /* key length */ 175 __u32 __tcpm_pad2; /* zero */ 176 __u8 tcpm_key[TCP_MD5SIG_MAXKEYLEN]; /* key (binary) */ 177 }; 178 179 /* for TCP_COOKIE_TRANSACTIONS (TCPCT) socket option */ 180 #define TCP_COOKIE_MIN 8 /* 64-bits */ 181 #define TCP_COOKIE_MAX 16 /* 128-bits */ 182 #define TCP_COOKIE_PAIR_SIZE (2*TCP_COOKIE_MAX) 183 184 /* Flags for both getsockopt and setsockopt */ 185 #define TCP_COOKIE_IN_ALWAYS (1 << 0) /* Discard SYN without cookie */ 186 #define TCP_COOKIE_OUT_NEVER (1 << 1) /* Prohibit outgoing cookies, 187 * supercedes everything. */ 188 189 /* Flags for getsockopt */ 190 #define TCP_S_DATA_IN (1 << 2) /* Was data received? */ 191 #define TCP_S_DATA_OUT (1 << 3) /* Was data sent? */ 192 193 /* TCP_COOKIE_TRANSACTIONS data */ 194 struct tcp_cookie_transactions { 195 __u16 tcpct_flags; /* see above */ 196 __u8 __tcpct_pad1; /* zero */ 197 __u8 tcpct_cookie_desired; /* bytes */ 198 __u16 tcpct_s_data_desired; /* bytes of variable data */ 199 __u16 tcpct_used; /* bytes in value */ 200 __u8 tcpct_value[TCP_MSS_DEFAULT]; 201 }; 202 203 #ifdef __KERNEL__ 204 205 #include <linux/skbuff.h> 206 #include <linux/dmaengine.h> 207 #include <net/sock.h> 208 #include <net/inet_connection_sock.h> 209 #include <net/inet_timewait_sock.h> 210 211 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb) 212 { 213 return (struct tcphdr *)skb_transport_header(skb); 214 } 215 216 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb) 217 { 218 return tcp_hdr(skb)->doff * 4; 219 } 220 221 static inline unsigned int tcp_optlen(const struct sk_buff *skb) 222 { 223 return (tcp_hdr(skb)->doff - 5) * 4; 224 } 225 226 /* This defines a selective acknowledgement block. */ 227 struct tcp_sack_block_wire { 228 __be32 start_seq; 229 __be32 end_seq; 230 }; 231 232 struct tcp_sack_block { 233 u32 start_seq; 234 u32 end_seq; 235 }; 236 237 struct tcp_options_received { 238 /* PAWS/RTTM data */ 239 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */ 240 u32 ts_recent; /* Time stamp to echo next */ 241 u32 rcv_tsval; /* Time stamp value */ 242 u32 rcv_tsecr; /* Time stamp echo reply */ 243 u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */ 244 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */ 245 dsack : 1, /* D-SACK is scheduled */ 246 wscale_ok : 1, /* Wscale seen on SYN packet */ 247 sack_ok : 4, /* SACK seen on SYN packet */ 248 snd_wscale : 4, /* Window scaling received from sender */ 249 rcv_wscale : 4; /* Window scaling to send to receiver */ 250 u8 cookie_plus:6, /* bytes in authenticator/cookie option */ 251 cookie_out_never:1, 252 cookie_in_always:1; 253 u8 num_sacks; /* Number of SACK blocks */ 254 u16 user_mss; /* mss requested by user in ioctl */ 255 u16 mss_clamp; /* Maximal mss, negotiated at connection setup */ 256 }; 257 258 static inline void tcp_clear_options(struct tcp_options_received *rx_opt) 259 { 260 rx_opt->tstamp_ok = rx_opt->sack_ok = 0; 261 rx_opt->wscale_ok = rx_opt->snd_wscale = 0; 262 rx_opt->cookie_plus = 0; 263 } 264 265 /* This is the max number of SACKS that we'll generate and process. It's safe 266 * to increase this, although since: 267 * size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8) 268 * only four options will fit in a standard TCP header */ 269 #define TCP_NUM_SACKS 4 270 271 struct tcp_cookie_values; 272 struct tcp_request_sock_ops; 273 274 struct tcp_request_sock { 275 struct inet_request_sock req; 276 #ifdef CONFIG_TCP_MD5SIG 277 /* Only used by TCP MD5 Signature so far. */ 278 const struct tcp_request_sock_ops *af_specific; 279 #endif 280 u32 rcv_isn; 281 u32 snt_isn; 282 }; 283 284 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req) 285 { 286 return (struct tcp_request_sock *)req; 287 } 288 289 struct tcp_sock { 290 /* inet_connection_sock has to be the first member of tcp_sock */ 291 struct inet_connection_sock inet_conn; 292 u16 tcp_header_len; /* Bytes of tcp header to send */ 293 u16 xmit_size_goal_segs; /* Goal for segmenting output packets */ 294 295 /* 296 * Header prediction flags 297 * 0x5?10 << 16 + snd_wnd in net byte order 298 */ 299 __be32 pred_flags; 300 301 /* 302 * RFC793 variables by their proper names. This means you can 303 * read the code and the spec side by side (and laugh ...) 304 * See RFC793 and RFC1122. The RFC writes these in capitals. 305 */ 306 u32 rcv_nxt; /* What we want to receive next */ 307 u32 copied_seq; /* Head of yet unread data */ 308 u32 rcv_wup; /* rcv_nxt on last window update sent */ 309 u32 snd_nxt; /* Next sequence we send */ 310 311 u32 snd_una; /* First byte we want an ack for */ 312 u32 snd_sml; /* Last byte of the most recently transmitted small packet */ 313 u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ 314 u32 lsndtime; /* timestamp of last sent data packet (for restart window) */ 315 316 /* Data for direct copy to user */ 317 struct { 318 struct sk_buff_head prequeue; 319 struct task_struct *task; 320 struct iovec *iov; 321 int memory; 322 int len; 323 #ifdef CONFIG_NET_DMA 324 /* members for async copy */ 325 struct dma_chan *dma_chan; 326 int wakeup; 327 struct dma_pinned_list *pinned_list; 328 dma_cookie_t dma_cookie; 329 #endif 330 } ucopy; 331 332 u32 snd_wl1; /* Sequence for window update */ 333 u32 snd_wnd; /* The window we expect to receive */ 334 u32 max_window; /* Maximal window ever seen from peer */ 335 u32 mss_cache; /* Cached effective mss, not including SACKS */ 336 337 u32 window_clamp; /* Maximal window to advertise */ 338 u32 rcv_ssthresh; /* Current window clamp */ 339 340 u32 frto_highmark; /* snd_nxt when RTO occurred */ 341 u16 advmss; /* Advertised MSS */ 342 u8 frto_counter; /* Number of new acks after RTO */ 343 u8 nonagle; /* Disable Nagle algorithm? */ 344 345 /* RTT measurement */ 346 u32 srtt; /* smoothed round trip time << 3 */ 347 u32 mdev; /* medium deviation */ 348 u32 mdev_max; /* maximal mdev for the last rtt period */ 349 u32 rttvar; /* smoothed mdev_max */ 350 u32 rtt_seq; /* sequence number to update rttvar */ 351 352 u32 packets_out; /* Packets which are "in flight" */ 353 u32 retrans_out; /* Retransmitted packets out */ 354 355 u16 urg_data; /* Saved octet of OOB data and control flags */ 356 u8 ecn_flags; /* ECN status bits. */ 357 u8 reordering; /* Packet reordering metric. */ 358 u32 snd_up; /* Urgent pointer */ 359 360 u8 keepalive_probes; /* num of allowed keep alive probes */ 361 /* 362 * Options received (usually on last packet, some only on SYN packets). 363 */ 364 struct tcp_options_received rx_opt; 365 366 /* 367 * Slow start and congestion control (see also Nagle, and Karn & Partridge) 368 */ 369 u32 snd_ssthresh; /* Slow start size threshold */ 370 u32 snd_cwnd; /* Sending congestion window */ 371 u32 snd_cwnd_cnt; /* Linear increase counter */ 372 u32 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */ 373 u32 snd_cwnd_used; 374 u32 snd_cwnd_stamp; 375 376 u32 rcv_wnd; /* Current receiver window */ 377 u32 write_seq; /* Tail(+1) of data held in tcp send buffer */ 378 u32 pushed_seq; /* Last pushed seq, required to talk to windows */ 379 u32 lost_out; /* Lost packets */ 380 u32 sacked_out; /* SACK'd packets */ 381 u32 fackets_out; /* FACK'd packets */ 382 u32 tso_deferred; 383 u32 bytes_acked; /* Appropriate Byte Counting - RFC3465 */ 384 385 /* from STCP, retrans queue hinting */ 386 struct sk_buff* lost_skb_hint; 387 struct sk_buff *scoreboard_skb_hint; 388 struct sk_buff *retransmit_skb_hint; 389 390 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */ 391 392 /* SACKs data, these 2 need to be together (see tcp_build_and_update_options) */ 393 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */ 394 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/ 395 396 struct tcp_sack_block recv_sack_cache[4]; 397 398 struct sk_buff *highest_sack; /* highest skb with SACK received 399 * (validity guaranteed only if 400 * sacked_out > 0) 401 */ 402 403 int lost_cnt_hint; 404 u32 retransmit_high; /* L-bits may be on up to this seqno */ 405 406 u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */ 407 408 u32 prior_ssthresh; /* ssthresh saved at recovery start */ 409 u32 high_seq; /* snd_nxt at onset of congestion */ 410 411 u32 retrans_stamp; /* Timestamp of the last retransmit, 412 * also used in SYN-SENT to remember stamp of 413 * the first SYN. */ 414 u32 undo_marker; /* tracking retrans started here. */ 415 int undo_retrans; /* number of undoable retransmissions. */ 416 u32 total_retrans; /* Total retransmits for entire connection */ 417 418 u32 urg_seq; /* Seq of received urgent pointer */ 419 unsigned int keepalive_time; /* time before keep alive takes place */ 420 unsigned int keepalive_intvl; /* time interval between keep alive probes */ 421 422 int linger2; 423 424 /* Receiver side RTT estimation */ 425 struct { 426 u32 rtt; 427 u32 seq; 428 u32 time; 429 } rcv_rtt_est; 430 431 /* Receiver queue space */ 432 struct { 433 int space; 434 u32 seq; 435 u32 time; 436 } rcvq_space; 437 438 /* TCP-specific MTU probe information. */ 439 struct { 440 u32 probe_seq_start; 441 u32 probe_seq_end; 442 } mtu_probe; 443 444 #ifdef CONFIG_TCP_MD5SIG 445 /* TCP AF-Specific parts; only used by MD5 Signature support so far */ 446 const struct tcp_sock_af_ops *af_specific; 447 448 /* TCP MD5 Signature Option information */ 449 struct tcp_md5sig_info *md5sig_info; 450 #endif 451 452 /* When the cookie options are generated and exchanged, then this 453 * object holds a reference to them (cookie_values->kref). Also 454 * contains related tcp_cookie_transactions fields. 455 */ 456 struct tcp_cookie_values *cookie_values; 457 }; 458 459 static inline struct tcp_sock *tcp_sk(const struct sock *sk) 460 { 461 return (struct tcp_sock *)sk; 462 } 463 464 struct tcp_timewait_sock { 465 struct inet_timewait_sock tw_sk; 466 u32 tw_rcv_nxt; 467 u32 tw_snd_nxt; 468 u32 tw_rcv_wnd; 469 u32 tw_ts_recent; 470 long tw_ts_recent_stamp; 471 #ifdef CONFIG_TCP_MD5SIG 472 u16 tw_md5_keylen; 473 u8 tw_md5_key[TCP_MD5SIG_MAXKEYLEN]; 474 #endif 475 /* Few sockets in timewait have cookies; in that case, then this 476 * object holds a reference to them (tw_cookie_values->kref). 477 */ 478 struct tcp_cookie_values *tw_cookie_values; 479 }; 480 481 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk) 482 { 483 return (struct tcp_timewait_sock *)sk; 484 } 485 486 #endif /* __KERNEL__ */ 487 488 #endif /* _LINUX_TCP_H */ 489