include/linux/tcp.h

   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, <waltje@uWalt.NL.Mugnet.ORG>
  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
  23 struct tcphdr {
  24         __u16   source;
  25         __u16   dest;
  26         __u32   seq;
  27         __u32   ack_seq;
  28 #if defined(__LITTLE_ENDIAN_BITFIELD)
  29         __u16   res1:4,
  30                 doff:4,
  31                 fin:1,
  32                 syn:1,
  33                 rst:1,
  34                 psh:1,
  35                 ack:1,
  36                 urg:1,
  37                 ece:1,
  38                 cwr:1;
  39 #elif defined(__BIG_ENDIAN_BITFIELD)
  40         __u16   doff:4,
  41                 res1:4,
  42                 cwr:1,
  43                 ece:1,
  44                 urg:1,
  45                 ack:1,
  46                 psh:1,
  47                 rst:1,
  48                 syn:1,
  49                 fin:1;
  50 #else
  51 #error  "Adjust your <asm/byteorder.h> defines"
  52 #endif
  53         __u16   window;
  54         __u16   check;
  55         __u16   urg_ptr;
  56 };
  57
  58
  59 enum {
  60   TCP_ESTABLISHED = 1,
  61   TCP_SYN_SENT,
  62   TCP_SYN_RECV,
  63   TCP_FIN_WAIT1,
  64   TCP_FIN_WAIT2,
  65   TCP_TIME_WAIT,
  66   TCP_CLOSE,
  67   TCP_CLOSE_WAIT,
  68   TCP_LAST_ACK,
  69   TCP_LISTEN,
  70   TCP_CLOSING,   /* now a valid state */
  71
  72   TCP_MAX_STATES /* Leave at the end! */
  73 };
  74
  75 #define TCP_STATE_MASK  0xF
  76 #define TCP_ACTION_FIN  (1 << 7)
  77
  78 enum {
  79   TCPF_ESTABLISHED = (1 << 1),
  80   TCPF_SYN_SENT  = (1 << 2),
  81   TCPF_SYN_RECV  = (1 << 3),
  82   TCPF_FIN_WAIT1 = (1 << 4),
  83   TCPF_FIN_WAIT2 = (1 << 5),
  84   TCPF_TIME_WAIT = (1 << 6),
  85   TCPF_CLOSE     = (1 << 7),
  86   TCPF_CLOSE_WAIT = (1 << 8),
  87   TCPF_LAST_ACK  = (1 << 9),
  88   TCPF_LISTEN    = (1 << 10),
  89   TCPF_CLOSING   = (1 << 11)
  90 };
  91
  92 /*
  93  *      The union cast uses a gcc extension to avoid aliasing problems
  94  *  (union is compatible to any of its members)
  95  *  This means this part of the code is -fstrict-aliasing safe now.
  96  */
  97 union tcp_word_hdr {
  98         struct tcphdr hdr;
  99         __u32             words[5];
 100 };
 101
 102 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
 103
 104 enum {
 105         TCP_FLAG_CWR = __constant_htonl(0x00800000),
 106         TCP_FLAG_ECE = __constant_htonl(0x00400000),
 107         TCP_FLAG_URG = __constant_htonl(0x00200000),
 108         TCP_FLAG_ACK = __constant_htonl(0x00100000),
 109         TCP_FLAG_PSH = __constant_htonl(0x00080000),
 110         TCP_FLAG_RST = __constant_htonl(0x00040000),
 111         TCP_FLAG_SYN = __constant_htonl(0x00020000),
 112         TCP_FLAG_FIN = __constant_htonl(0x00010000),
 113         TCP_RESERVED_BITS = __constant_htonl(0x0F000000),
 114         TCP_DATA_OFFSET = __constant_htonl(0xF0000000)
 115 };
 116
 117 /* TCP socket options */
 118 #define TCP_NODELAY             1       /* Turn off Nagle's algorithm. */
 119 #define TCP_MAXSEG              2       /* Limit MSS */
 120 #define TCP_CORK                3       /* Never send partially complete segments */
 121 #define TCP_KEEPIDLE            4       /* Start keeplives after this period */
 122 #define TCP_KEEPINTVL           5       /* Interval between keepalives */
 123 #define TCP_KEEPCNT             6       /* Number of keepalives before death */
 124 #define TCP_SYNCNT              7       /* Number of SYN retransmits */
 125 #define TCP_LINGER2             8       /* Life time of orphaned FIN-WAIT-2 state */
 126 #define TCP_DEFER_ACCEPT        9       /* Wake up listener only when data arrive */
 127 #define TCP_WINDOW_CLAMP        10      /* Bound advertised window */
 128 #define TCP_INFO                11      /* Information about this connection. */
 129 #define TCP_QUICKACK            12      /* Block/reenable quick acks */
 130
 131 #define TCPI_OPT_TIMESTAMPS     1
 132 #define TCPI_OPT_SACK           2
 133 #define TCPI_OPT_WSCALE         4
 134 #define TCPI_OPT_ECN            8
 135
 136 enum tcp_ca_state
 137 {
 138         TCP_CA_Open = 0,
 139 #define TCPF_CA_Open    (1<<TCP_CA_Open)
 140         TCP_CA_Disorder = 1,
 141 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
 142         TCP_CA_CWR = 2,
 143 #define TCPF_CA_CWR     (1<<TCP_CA_CWR)
 144         TCP_CA_Recovery = 3,
 145 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
 146         TCP_CA_Loss = 4
 147 #define TCPF_CA_Loss    (1<<TCP_CA_Loss)
 148 };
 149
 150 struct tcp_info
 151 {
 152         __u8    tcpi_state;
 153         __u8    tcpi_ca_state;
 154         __u8    tcpi_retransmits;
 155         __u8    tcpi_probes;
 156         __u8    tcpi_backoff;
 157         __u8    tcpi_options;
 158         __u8    tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
 159
 160         __u32   tcpi_rto;
 161         __u32   tcpi_ato;
 162         __u32   tcpi_snd_mss;
 163         __u32   tcpi_rcv_mss;
 164
 165         __u32   tcpi_unacked;
 166         __u32   tcpi_sacked;
 167         __u32   tcpi_lost;
 168         __u32   tcpi_retrans;
 169         __u32   tcpi_fackets;
 170
 171         /* Times. */
 172         __u32   tcpi_last_data_sent;
 173         __u32   tcpi_last_ack_sent;     /* Not remembered, sorry. */
 174         __u32   tcpi_last_data_recv;
 175         __u32   tcpi_last_ack_recv;
 176
 177         /* Metrics. */
 178         __u32   tcpi_pmtu;
 179         __u32   tcpi_rcv_ssthresh;
 180         __u32   tcpi_rtt;
 181         __u32   tcpi_rttvar;
 182         __u32   tcpi_snd_ssthresh;
 183         __u32   tcpi_snd_cwnd;
 184         __u32   tcpi_advmss;
 185         __u32   tcpi_reordering;
 186
 187         __u32   tcpi_rcv_rtt;
 188         __u32   tcpi_rcv_space;
 189 };
 190
 191 #ifdef __KERNEL__
 192
 193 #include <linux/config.h>
 194 #include <linux/skbuff.h>
 195 #include <linux/ip.h>
 196 #include <net/sock.h>
 197
 198 /* This defines a selective acknowledgement block. */
 199 struct tcp_sack_block {
 200         __u32   start_seq;
 201         __u32   end_seq;
 202 };
 203
 204 typedef struct tcp_pcount {
 205         __u32   val;
 206 } tcp_pcount_t;
 207
 208 enum tcp_congestion_algo {
 209         TCP_RENO=0,
 210         TCP_VEGAS,
 211         TCP_WESTWOOD,
 212         TCP_BIC,
 213 };
 214
 215 struct tcp_opt {
 216         int     tcp_header_len; /* Bytes of tcp header to send          */
 217
 218 /*
 219  *      Header prediction flags
 220  *      0x5?10 << 16 + snd_wnd in net byte order
 221  */
 222         __u32   pred_flags;
 223
 224 /*
 225  *      RFC793 variables by their proper names. This means you can
 226  *      read the code and the spec side by side (and laugh ...)
 227  *      See RFC793 and RFC1122. The RFC writes these in capitals.
 228  */
 229         __u32   rcv_nxt;        /* What we want to receive next         */
 230         __u32   snd_nxt;        /* Next sequence we send                */
 231
 232         __u32   snd_una;        /* First byte we want an ack for        */
 233         __u32   snd_sml;        /* Last byte of the most recently transmitted small packet */
 234         __u32   rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 235         __u32   lsndtime;       /* timestamp of last sent data packet (for restart window) */
 236         struct tcp_bind_bucket *bind_hash;
 237         /* Delayed ACK control data */
 238         struct {
 239                 __u8    pending;        /* ACK is pending */
 240                 __u8    quick;          /* Scheduled number of quick acks       */
 241                 __u8    pingpong;       /* The session is interactive           */
 242                 __u8    blocked;        /* Delayed ACK was blocked by socket lock*/
 243                 __u32   ato;            /* Predicted tick of soft clock         */
 244                 unsigned long timeout;  /* Currently scheduled timeout          */
 245                 __u32   lrcvtime;       /* timestamp of last received data packet*/
 246                 __u16   last_seg_size;  /* Size of last incoming segment        */
 247                 __u16   rcv_mss;        /* MSS used for delayed ACK decisions   */
 248         } ack;
 249
 250         /* Data for direct copy to user */
 251         struct {
 252                 struct sk_buff_head     prequeue;
 253                 struct task_struct      *task;
 254                 struct iovec            *iov;
 255                 int                     memory;
 256                 int                     len;
 257         } ucopy;
 258
 259         __u32   snd_wl1;        /* Sequence for window update           */
 260         __u32   snd_wnd;        /* The window we expect to receive      */
 261         __u32   max_window;     /* Maximal window ever seen from peer   */
 262         __u32   pmtu_cookie;    /* Last pmtu seen by socket             */
 263         __u32   mss_cache;      /* Cached effective mss, not including SACKS */
 264         __u16   mss_cache_std;  /* Like mss_cache, but without TSO */
 265         __u16   mss_clamp;      /* Maximal mss, negotiated at connection setup */
 266         __u16   ext_header_len; /* Network protocol overhead (IP/IPv6 options) */
 267         __u16   ext2_header_len;/* Options depending on route */
 268         __u8    ca_state;       /* State of fast-retransmit machine     */
 269         __u8    retransmits;    /* Number of unrecovered RTO timeouts.  */
 270
 271         __u8    reordering;     /* Packet reordering metric.            */
 272         __u8    frto_counter;   /* Number of new acks after RTO */
 273         __u32   frto_highmark;  /* snd_nxt when RTO occurred */
 274
 275         __u8    adv_cong;       /* Using Vegas, Westwood, or BIC */
 276         __u8    defer_accept;   /* User waits for some data after accept() */
 277         /* one byte hole, try to pack */
 278
 279 /* RTT measurement */
 280         __u8    backoff;        /* backoff                              */
 281         __u32   srtt;           /* smoothed round trip time << 3        */
 282         __u32   mdev;           /* medium deviation                     */
 283         __u32   mdev_max;       /* maximal mdev for the last rtt period */
 284         __u32   rttvar;         /* smoothed mdev_max                    */
 285         __u32   rtt_seq;        /* sequence number to update rttvar     */
 286         __u32   rto;            /* retransmit timeout                   */
 287
 288         tcp_pcount_t packets_out; /* Packets which are "in flight"      */
 289         tcp_pcount_t left_out;    /* Packets which leaved network       */
 290         tcp_pcount_t retrans_out; /* Retransmitted packets out          */
 291
 292
 293 /*
 294  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 295  */
 296         __u32   snd_ssthresh;   /* Slow start size threshold            */
 297         __u32   snd_cwnd;       /* Sending congestion window            */
 298         __u16   snd_cwnd_cnt;   /* Linear increase counter              */
 299         __u16   snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 300         __u32   snd_cwnd_used;
 301         __u32   snd_cwnd_stamp;
 302
 303         /* Two commonly used timers in both sender and receiver paths. */
 304         unsigned long           timeout;
 305         struct timer_list       retransmit_timer;       /* Resend (no ack)      */
 306         struct timer_list       delack_timer;           /* Ack delay            */
 307
 308         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 309
 310         struct tcp_func         *af_specific;   /* Operations which are AF_INET{4,6} specific   */
 311
 312         __u32   rcv_wnd;        /* Current receiver window              */
 313         __u32   rcv_wup;        /* rcv_nxt on last window update sent   */
 314         __u32   write_seq;      /* Tail(+1) of data held in tcp send buffer */
 315         __u32   pushed_seq;     /* Last pushed seq, required to talk to windows */
 316         __u32   copied_seq;     /* Head of yet unread data              */
 317 /*
 318  *      Options received (usually on last packet, some only on SYN packets).
 319  */
 320         char    tstamp_ok,      /* TIMESTAMP seen on SYN packet         */
 321                 wscale_ok,      /* Wscale seen on SYN packet            */
 322                 sack_ok;        /* SACK seen on SYN packet              */
 323         char    saw_tstamp;     /* Saw TIMESTAMP on last packet         */
 324         __u8    snd_wscale;     /* Window scaling received from sender  */
 325         __u8    rcv_wscale;     /* Window scaling to send to receiver   */
 326         __u8    nonagle;        /* Disable Nagle algorithm?             */
 327         __u8    keepalive_probes; /* num of allowed keep alive probes   */
 328
 329 /*      PAWS/RTTM data  */
 330         __u32   rcv_tsval;      /* Time stamp value                     */
 331         __u32   rcv_tsecr;      /* Time stamp echo reply                */
 332         __u32   ts_recent;      /* Time stamp to echo next              */
 333         long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
 334
 335 /*      SACKs data      */
 336         __u16   user_mss;       /* mss requested by user in ioctl */
 337         __u8    dsack;          /* D-SACK is scheduled                  */
 338         __u8    eff_sacks;      /* Size of SACK array to send with next packet */
 339         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 340         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 341
 342         __u32   window_clamp;   /* Maximal window to advertise          */
 343         __u32   rcv_ssthresh;   /* Current window clamp                 */
 344         __u8    probes_out;     /* unanswered 0 window probes           */
 345         __u8    num_sacks;      /* Number of SACK blocks                */
 346         __u16   advmss;         /* Advertised MSS                       */
 347
 348         __u8    syn_retries;    /* num of allowed syn retries */
 349         __u8    ecn_flags;      /* ECN status bits.                     */
 350         __u16   prior_ssthresh; /* ssthresh saved at recovery start     */
 351         tcp_pcount_t lost_out;  /* Lost packets                 */
 352         tcp_pcount_t sacked_out;/* SACK'd packets                       */
 353         tcp_pcount_t fackets_out;/* FACK'd packets                      */
 354         __u32   high_seq;       /* snd_nxt at onset of congestion       */
 355
 356         __u32   retrans_stamp;  /* Timestamp of the last retransmit,
 357                                  * also used in SYN-SENT to remember stamp of
 358                                  * the first SYN. */
 359         __u32   undo_marker;    /* tracking retrans started here. */
 360         int     undo_retrans;   /* number of undoable retransmissions. */
 361         __u32   urg_seq;        /* Seq of received urgent pointer */
 362         __u16   urg_data;       /* Saved octet of OOB data and control flags */
 363         __u8    pending;        /* Scheduled timer event        */
 364         __u8    urg_mode;       /* In urgent mode               */
 365         __u32   snd_up;         /* Urgent pointer               */
 366
 367         /* The syn_wait_lock is necessary only to avoid proc interface having
 368          * to grab the main lock sock while browsing the listening hash
 369          * (otherwise it's deadlock prone).
 370          * This lock is acquired in read mode only from listening_get_next()
 371          * and it's acquired in write mode _only_ from code that is actively
 372          * changing the syn_wait_queue. All readers that are holding
 373          * the master sock lock don't need to grab this lock in read mode
 374          * too as the syn_wait_queue writes are always protected from
 375          * the main sock lock.
 376          */
 377         rwlock_t                syn_wait_lock;
 378         struct tcp_listen_opt   *listen_opt;
 379
 380         /* FIFO of established children */
 381         struct open_request     *accept_queue;
 382         struct open_request     *accept_queue_tail;
 383
 384         unsigned int            keepalive_time;   /* time before keep alive takes place */
 385         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 386         int                     linger2;
 387
 388         unsigned long last_synq_overflow;
 389
 390 /* Receiver side RTT estimation */
 391         struct {
 392                 __u32   rtt;
 393                 __u32   seq;
 394                 __u32   time;
 395         } rcv_rtt_est;
 396
 397 /* Receiver queue space */
 398         struct {
 399                 int     space;
 400                 __u32   seq;
 401                 __u32   time;
 402         } rcvq_space;
 403
 404 /* TCP Westwood structure */
 405         struct {
 406                 __u32    bw_ns_est;        /* first bandwidth estimation..not too smoothed 8) */
 407                 __u32    bw_est;           /* bandwidth estimate */
 408                 __u32    rtt_win_sx;       /* here starts a new evaluation... */
 409                 __u32    bk;
 410                 __u32    snd_una;          /* used for evaluating the number of acked bytes */
 411                 __u32    cumul_ack;
 412                 __u32    accounted;
 413                 __u32    rtt;
 414                 __u32    rtt_min;          /* minimum observed RTT */
 415         } westwood;
 416
 417 /* Vegas variables */
 418         struct {
 419                 __u32   beg_snd_nxt;    /* right edge during last RTT */
 420                 __u32   beg_snd_una;    /* left edge  during last RTT */
 421                 __u32   beg_snd_cwnd;   /* saves the size of the cwnd */
 422                 __u8    doing_vegas_now;/* if true, do vegas for this RTT */
 423                 __u16   cntRTT;         /* # of RTTs measured within last RTT */
 424                 __u32   minRTT;         /* min of RTTs measured within last RTT (in usec) */
 425                 __u32   baseRTT;        /* the min of all Vegas RTT measurements seen (in usec) */
 426         } vegas;
 427
 428         /* BI TCP Parameters */
 429         struct {
 430                 __u32   cnt;            /* increase cwnd by 1 after this number of ACKs */
 431                 __u32   last_max_cwnd;  /* last maximium snd_cwnd */
 432                 __u32   last_cwnd;      /* the last snd_cwnd */
 433                 __u32   last_stamp;     /* time when updated last_cwnd */
 434         } bictcp;
 435 };
 436
 437 /* WARNING: don't change the layout of the members in tcp_sock! */
 438 struct tcp_sock {
 439         struct sock       sk;
 440 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 441         struct ipv6_pinfo *pinet6;
 442 #endif
 443         struct inet_opt   inet;
 444         struct tcp_opt    tcp;
 445 };
 446
 447 static inline struct tcp_opt * tcp_sk(const struct sock *__sk)
 448 {
 449         return &((struct tcp_sock *)__sk)->tcp;
 450 }
 451
 452 #endif
 453
 454 #endif  /* _LINUX_TCP_H */