LXR linux/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#include <linux/socket.h>
  23
  24struct 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 */
  64union 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
  71enum { 
  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#define TCP_THIN_LINEAR_TIMEOUTS 16      /* Use linear timeouts for thin streams*/
 107#define TCP_THIN_DUPACK         17      /* Fast retrans. after 1 dupack */
 108
 109/* for TCP_INFO socket option */
 110#define TCPI_OPT_TIMESTAMPS     1
 111#define TCPI_OPT_SACK           2
 112#define TCPI_OPT_WSCALE         4
 113#define TCPI_OPT_ECN            8
 114
 115enum tcp_ca_state {
 116        TCP_CA_Open = 0,
 117#define TCPF_CA_Open    (1<<TCP_CA_Open)
 118        TCP_CA_Disorder = 1,
 119#define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
 120        TCP_CA_CWR = 2,
 121#define TCPF_CA_CWR     (1<<TCP_CA_CWR)
 122        TCP_CA_Recovery = 3,
 123#define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
 124        TCP_CA_Loss = 4
 125#define TCPF_CA_Loss    (1<<TCP_CA_Loss)
 126};
 127
 128struct tcp_info {
 129        __u8    tcpi_state;
 130        __u8    tcpi_ca_state;
 131        __u8    tcpi_retransmits;
 132        __u8    tcpi_probes;
 133        __u8    tcpi_backoff;
 134        __u8    tcpi_options;
 135        __u8    tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
 136
 137        __u32   tcpi_rto;
 138        __u32   tcpi_ato;
 139        __u32   tcpi_snd_mss;
 140        __u32   tcpi_rcv_mss;
 141
 142        __u32   tcpi_unacked;
 143        __u32   tcpi_sacked;
 144        __u32   tcpi_lost;
 145        __u32   tcpi_retrans;
 146        __u32   tcpi_fackets;
 147
 148        /* Times. */
 149        __u32   tcpi_last_data_sent;
 150        __u32   tcpi_last_ack_sent;     /* Not remembered, sorry. */
 151        __u32   tcpi_last_data_recv;
 152        __u32   tcpi_last_ack_recv;
 153
 154        /* Metrics. */
 155        __u32   tcpi_pmtu;
 156        __u32   tcpi_rcv_ssthresh;
 157        __u32   tcpi_rtt;
 158        __u32   tcpi_rttvar;
 159        __u32   tcpi_snd_ssthresh;
 160        __u32   tcpi_snd_cwnd;
 161        __u32   tcpi_advmss;
 162        __u32   tcpi_reordering;
 163
 164        __u32   tcpi_rcv_rtt;
 165        __u32   tcpi_rcv_space;
 166
 167        __u32   tcpi_total_retrans;
 168};
 169
 170/* for TCP_MD5SIG socket option */
 171#define TCP_MD5SIG_MAXKEYLEN    80
 172
 173struct tcp_md5sig {
 174        struct __kernel_sockaddr_storage tcpm_addr;     /* address associated */
 175        __u16   __tcpm_pad1;                            /* zero */
 176        __u16   tcpm_keylen;                            /* key length */
 177        __u32   __tcpm_pad2;                            /* zero */
 178        __u8    tcpm_key[TCP_MD5SIG_MAXKEYLEN];         /* key (binary) */
 179};
 180
 181/* for TCP_COOKIE_TRANSACTIONS (TCPCT) socket option */
 182#define TCP_COOKIE_MIN           8              /*  64-bits */
 183#define TCP_COOKIE_MAX          16              /* 128-bits */
 184#define TCP_COOKIE_PAIR_SIZE    (2*TCP_COOKIE_MAX)
 185
 186/* Flags for both getsockopt and setsockopt */
 187#define TCP_COOKIE_IN_ALWAYS    (1 << 0)        /* Discard SYN without cookie */
 188#define TCP_COOKIE_OUT_NEVER    (1 << 1)        /* Prohibit outgoing cookies,
 189                                                 * supercedes everything. */
 190
 191/* Flags for getsockopt */
 192#define TCP_S_DATA_IN           (1 << 2)        /* Was data received? */
 193#define TCP_S_DATA_OUT          (1 << 3)        /* Was data sent? */
 194
 195/* TCP_COOKIE_TRANSACTIONS data */
 196struct tcp_cookie_transactions {
 197        __u16   tcpct_flags;                    /* see above */
 198        __u8    __tcpct_pad1;                   /* zero */
 199        __u8    tcpct_cookie_desired;           /* bytes */
 200        __u16   tcpct_s_data_desired;           /* bytes of variable data */
 201        __u16   tcpct_used;                     /* bytes in value */
 202        __u8    tcpct_value[TCP_MSS_DEFAULT];
 203};
 204
 205#ifdef __KERNEL__
 206
 207#include <linux/skbuff.h>
 208#include <linux/dmaengine.h>
 209#include <net/sock.h>
 210#include <net/inet_connection_sock.h>
 211#include <net/inet_timewait_sock.h>
 212
 213static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
 214{
 215        return (struct tcphdr *)skb_transport_header(skb);
 216}
 217
 218static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
 219{
 220        return tcp_hdr(skb)->doff * 4;
 221}
 222
 223static inline unsigned int tcp_optlen(const struct sk_buff *skb)
 224{
 225        return (tcp_hdr(skb)->doff - 5) * 4;
 226}
 227
 228/* This defines a selective acknowledgement block. */
 229struct tcp_sack_block_wire {
 230        __be32  start_seq;
 231        __be32  end_seq;
 232};
 233
 234struct tcp_sack_block {
 235        u32     start_seq;
 236        u32     end_seq;
 237};
 238
 239struct tcp_options_received {
 240/*      PAWS/RTTM data  */
 241        long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
 242        u32     ts_recent;      /* Time stamp to echo next              */
 243        u32     rcv_tsval;      /* Time stamp value                     */
 244        u32     rcv_tsecr;      /* Time stamp echo reply                */
 245        u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
 246                tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
 247                dsack : 1,      /* D-SACK is scheduled                  */
 248                wscale_ok : 1,  /* Wscale seen on SYN packet            */
 249                sack_ok : 4,    /* SACK seen on SYN packet              */
 250                snd_wscale : 4, /* Window scaling received from sender  */
 251                rcv_wscale : 4; /* Window scaling to send to receiver   */
 252        u8      cookie_plus:6,  /* bytes in authenticator/cookie option */
 253                cookie_out_never:1,
 254                cookie_in_always:1;
 255        u8      num_sacks;      /* Number of SACK blocks                */
 256        u16     user_mss;       /* mss requested by user in ioctl       */
 257        u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
 258};
 259
 260static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
 261{
 262        rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
 263        rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
 264        rx_opt->cookie_plus = 0;
 265}
 266
 267/* This is the max number of SACKS that we'll generate and process. It's safe
 268 * to increase this, although since:
 269 *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
 270 * only four options will fit in a standard TCP header */
 271#define TCP_NUM_SACKS 4
 272
 273struct tcp_cookie_values;
 274struct tcp_request_sock_ops;
 275
 276struct tcp_request_sock {
 277        struct inet_request_sock        req;
 278#ifdef CONFIG_TCP_MD5SIG
 279        /* Only used by TCP MD5 Signature so far. */
 280        const struct tcp_request_sock_ops *af_specific;
 281#endif
 282        u32                             rcv_isn;
 283        u32                             snt_isn;
 284};
 285
 286static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 287{
 288        return (struct tcp_request_sock *)req;
 289}
 290
 291struct tcp_sock {
 292        /* inet_connection_sock has to be the first member of tcp_sock */
 293        struct inet_connection_sock     inet_conn;
 294        u16     tcp_header_len; /* Bytes of tcp header to send          */
 295        u16     xmit_size_goal_segs; /* Goal for segmenting output packets */
 296
 297/*
 298 *      Header prediction flags
 299 *      0x5?10 << 16 + snd_wnd in net byte order
 300 */
 301        __be32  pred_flags;
 302
 303/*
 304 *      RFC793 variables by their proper names. This means you can
 305 *      read the code and the spec side by side (and laugh ...)
 306 *      See RFC793 and RFC1122. The RFC writes these in capitals.
 307 */
 308        u32     rcv_nxt;        /* What we want to receive next         */
 309        u32     copied_seq;     /* Head of yet unread data              */
 310        u32     rcv_wup;        /* rcv_nxt on last window update sent   */
 311        u32     snd_nxt;        /* Next sequence we send                */
 312
 313        u32     snd_una;        /* First byte we want an ack for        */
 314        u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
 315        u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 316        u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 317
 318        /* Data for direct copy to user */
 319        struct {
 320                struct sk_buff_head     prequeue;
 321                struct task_struct      *task;
 322                struct iovec            *iov;
 323                int                     memory;
 324                int                     len;
 325#ifdef CONFIG_NET_DMA
 326                /* members for async copy */
 327                struct dma_chan         *dma_chan;
 328                int                     wakeup;
 329                struct dma_pinned_list  *pinned_list;
 330                dma_cookie_t            dma_cookie;
 331#endif
 332        } ucopy;
 333
 334        u32     snd_wl1;        /* Sequence for window update           */
 335        u32     snd_wnd;        /* The window we expect to receive      */
 336        u32     max_window;     /* Maximal window ever seen from peer   */
 337        u32     mss_cache;      /* Cached effective mss, not including SACKS */
 338
 339        u32     window_clamp;   /* Maximal window to advertise          */
 340        u32     rcv_ssthresh;   /* Current window clamp                 */
 341
 342        u32     frto_highmark;  /* snd_nxt when RTO occurred */
 343        u16     advmss;         /* Advertised MSS                       */
 344        u8      frto_counter;   /* Number of new acks after RTO */
 345        u8      nonagle     : 4,/* Disable Nagle algorithm?             */
 346                thin_lto    : 1,/* Use linear timeouts for thin streams */
 347                thin_dupack : 1,/* Fast retransmit on first dupack      */
 348                unused      : 2;
 349
 350/* RTT measurement */
 351        u32     srtt;           /* smoothed round trip time << 3        */
 352        u32     mdev;           /* medium deviation                     */
 353        u32     mdev_max;       /* maximal mdev for the last rtt period */
 354        u32     rttvar;         /* smoothed mdev_max                    */
 355        u32     rtt_seq;        /* sequence number to update rttvar     */
 356
 357        u32     packets_out;    /* Packets which are "in flight"        */
 358        u32     retrans_out;    /* Retransmitted packets out            */
 359
 360        u16     urg_data;       /* Saved octet of OOB data and control flags */
 361        u8      ecn_flags;      /* ECN status bits.                     */
 362        u8      reordering;     /* Packet reordering metric.            */
 363        u32     snd_up;         /* Urgent pointer               */
 364
 365        u8      keepalive_probes; /* num of allowed keep alive probes   */
 366/*
 367 *      Options received (usually on last packet, some only on SYN packets).
 368 */
 369        struct tcp_options_received rx_opt;
 370
 371/*
 372 *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 373 */
 374        u32     snd_ssthresh;   /* Slow start size threshold            */
 375        u32     snd_cwnd;       /* Sending congestion window            */
 376        u32     snd_cwnd_cnt;   /* Linear increase counter              */
 377        u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 378        u32     snd_cwnd_used;
 379        u32     snd_cwnd_stamp;
 380
 381        u32     rcv_wnd;        /* Current receiver window              */
 382        u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 383        u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 384        u32     lost_out;       /* Lost packets                 */
 385        u32     sacked_out;     /* SACK'd packets                       */
 386        u32     fackets_out;    /* FACK'd packets                       */
 387        u32     tso_deferred;
 388        u32     bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
 389
 390        /* from STCP, retrans queue hinting */
 391        struct sk_buff* lost_skb_hint;
 392        struct sk_buff *scoreboard_skb_hint;
 393        struct sk_buff *retransmit_skb_hint;
 394
 395        struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 396
 397        /* SACKs data, these 2 need to be together (see tcp_build_and_update_options) */
 398        struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 399        struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 400
 401        struct tcp_sack_block recv_sack_cache[4];
 402
 403        struct sk_buff *highest_sack;   /* highest skb with SACK received
 404                                         * (validity guaranteed only if
 405                                         * sacked_out > 0)
 406                                         */
 407
 408        int     lost_cnt_hint;
 409        u32     retransmit_high;        /* L-bits may be on up to this seqno */
 410
 411        u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 412
 413        u32     prior_ssthresh; /* ssthresh saved at recovery start     */
 414        u32     high_seq;       /* snd_nxt at onset of congestion       */
 415
 416        u32     retrans_stamp;  /* Timestamp of the last retransmit,
 417                                 * also used in SYN-SENT to remember stamp of
 418                                 * the first SYN. */
 419        u32     undo_marker;    /* tracking retrans started here. */
 420        int     undo_retrans;   /* number of undoable retransmissions. */
 421        u32     total_retrans;  /* Total retransmits for entire connection */
 422
 423        u32     urg_seq;        /* Seq of received urgent pointer */
 424        unsigned int            keepalive_time;   /* time before keep alive takes place */
 425        unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 426
 427        int                     linger2;
 428
 429/* Receiver side RTT estimation */
 430        struct {
 431                u32     rtt;
 432                u32     seq;
 433                u32     time;
 434        } rcv_rtt_est;
 435
 436/* Receiver queue space */
 437        struct {
 438                int     space;
 439                u32     seq;
 440                u32     time;
 441        } rcvq_space;
 442
 443/* TCP-specific MTU probe information. */
 444        struct {
 445                u32               probe_seq_start;
 446                u32               probe_seq_end;
 447        } mtu_probe;
 448
 449#ifdef CONFIG_TCP_MD5SIG
 450/* TCP AF-Specific parts; only used by MD5 Signature support so far */
 451        const struct tcp_sock_af_ops    *af_specific;
 452
 453/* TCP MD5 Signature Option information */
 454        struct tcp_md5sig_info  *md5sig_info;
 455#endif
 456
 457        /* When the cookie options are generated and exchanged, then this
 458         * object holds a reference to them (cookie_values->kref).  Also
 459         * contains related tcp_cookie_transactions fields.
 460         */
 461        struct tcp_cookie_values  *cookie_values;
 462};
 463
 464static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 465{
 466        return (struct tcp_sock *)sk;
 467}
 468
 469struct tcp_timewait_sock {
 470        struct inet_timewait_sock tw_sk;
 471        u32                       tw_rcv_nxt;
 472        u32                       tw_snd_nxt;
 473        u32                       tw_rcv_wnd;
 474        u32                       tw_ts_recent;
 475        long                      tw_ts_recent_stamp;
 476#ifdef CONFIG_TCP_MD5SIG
 477        u16                       tw_md5_keylen;
 478        u8                        tw_md5_key[TCP_MD5SIG_MAXKEYLEN];
 479#endif
 480        /* Few sockets in timewait have cookies; in that case, then this
 481         * object holds a reference to them (tw_cookie_values->kref).
 482         */
 483        struct tcp_cookie_values  *tw_cookie_values;
 484};
 485
 486static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 487{
 488        return (struct tcp_timewait_sock *)sk;
 489}
 490
 491#endif  /* __KERNEL__ */
 492
 493#endif  /* _LINUX_TCP_H */
 494