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
  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_htonl(0x00800000),
  73         TCP_FLAG_ECE = __constant_htonl(0x00400000),
  74         TCP_FLAG_URG = __constant_htonl(0x00200000),
  75         TCP_FLAG_ACK = __constant_htonl(0x00100000),
  76         TCP_FLAG_PSH = __constant_htonl(0x00080000),
  77         TCP_FLAG_RST = __constant_htonl(0x00040000),
  78         TCP_FLAG_SYN = __constant_htonl(0x00020000),
  79         TCP_FLAG_FIN = __constant_htonl(0x00010000),
  80         TCP_RESERVED_BITS = __constant_htonl(0x0F000000),
  81         TCP_DATA_OFFSET = __constant_htonl(0xF0000000)
  82 };
  83
  84 /* TCP socket options */
  85 #define TCP_NODELAY             1       /* Turn off Nagle's algorithm. */
  86 #define TCP_MAXSEG              2       /* Limit MSS */
  87 #define TCP_CORK                3       /* Never send partially complete segments */
  88 #define TCP_KEEPIDLE            4       /* Start keeplives after this period */
  89 #define TCP_KEEPINTVL           5       /* Interval between keepalives */
  90 #define TCP_KEEPCNT             6       /* Number of keepalives before death */
  91 #define TCP_SYNCNT              7       /* Number of SYN retransmits */
  92 #define TCP_LINGER2             8       /* Life time of orphaned FIN-WAIT-2 state */
  93 #define TCP_DEFER_ACCEPT        9       /* Wake up listener only when data arrive */
  94 #define TCP_WINDOW_CLAMP        10      /* Bound advertised window */
  95 #define TCP_INFO                11      /* Information about this connection. */
  96 #define TCP_QUICKACK            12      /* Block/reenable quick acks */
  97 #define TCP_CONGESTION          13      /* Congestion control algorithm */
  98 #define TCP_MD5SIG              14      /* TCP MD5 Signature (RFC2385) */
  99
 100 #define TCPI_OPT_TIMESTAMPS     1
 101 #define TCPI_OPT_SACK           2
 102 #define TCPI_OPT_WSCALE         4
 103 #define TCPI_OPT_ECN            8
 104
 105 enum tcp_ca_state
 106 {
 107         TCP_CA_Open = 0,
 108 #define TCPF_CA_Open    (1<<TCP_CA_Open)
 109         TCP_CA_Disorder = 1,
 110 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
 111         TCP_CA_CWR = 2,
 112 #define TCPF_CA_CWR     (1<<TCP_CA_CWR)
 113         TCP_CA_Recovery = 3,
 114 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
 115         TCP_CA_Loss = 4
 116 #define TCPF_CA_Loss    (1<<TCP_CA_Loss)
 117 };
 118
 119 struct tcp_info
 120 {
 121         __u8    tcpi_state;
 122         __u8    tcpi_ca_state;
 123         __u8    tcpi_retransmits;
 124         __u8    tcpi_probes;
 125         __u8    tcpi_backoff;
 126         __u8    tcpi_options;
 127         __u8    tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
 128
 129         __u32   tcpi_rto;
 130         __u32   tcpi_ato;
 131         __u32   tcpi_snd_mss;
 132         __u32   tcpi_rcv_mss;
 133
 134         __u32   tcpi_unacked;
 135         __u32   tcpi_sacked;
 136         __u32   tcpi_lost;
 137         __u32   tcpi_retrans;
 138         __u32   tcpi_fackets;
 139
 140         /* Times. */
 141         __u32   tcpi_last_data_sent;
 142         __u32   tcpi_last_ack_sent;     /* Not remembered, sorry. */
 143         __u32   tcpi_last_data_recv;
 144         __u32   tcpi_last_ack_recv;
 145
 146         /* Metrics. */
 147         __u32   tcpi_pmtu;
 148         __u32   tcpi_rcv_ssthresh;
 149         __u32   tcpi_rtt;
 150         __u32   tcpi_rttvar;
 151         __u32   tcpi_snd_ssthresh;
 152         __u32   tcpi_snd_cwnd;
 153         __u32   tcpi_advmss;
 154         __u32   tcpi_reordering;
 155
 156         __u32   tcpi_rcv_rtt;
 157         __u32   tcpi_rcv_space;
 158
 159         __u32   tcpi_total_retrans;
 160 };
 161
 162 /* for TCP_MD5SIG socket option */
 163 #define TCP_MD5SIG_MAXKEYLEN    80
 164
 165 struct tcp_md5sig {
 166         struct __kernel_sockaddr_storage tcpm_addr;     /* address associated */
 167         __u16   __tcpm_pad1;                            /* zero */
 168         __u16   tcpm_keylen;                            /* key length */
 169         __u32   __tcpm_pad2;                            /* zero */
 170         __u8    tcpm_key[TCP_MD5SIG_MAXKEYLEN];         /* key (binary) */
 171 };
 172
 173 #ifdef __KERNEL__
 174
 175 #include <linux/skbuff.h>
 176 #include <linux/dmaengine.h>
 177 #include <net/sock.h>
 178 #include <net/inet_connection_sock.h>
 179 #include <net/inet_timewait_sock.h>
 180
 181 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
 182 {
 183         return skb->h.th->doff * 4;
 184 }
 185
 186 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
 187 {
 188         return (skb->h.th->doff - 5) * 4;
 189 }
 190
 191 /* This defines a selective acknowledgement block. */
 192 struct tcp_sack_block_wire {
 193         __be32  start_seq;
 194         __be32  end_seq;
 195 };
 196
 197 struct tcp_sack_block {
 198         u32     start_seq;
 199         u32     end_seq;
 200 };
 201
 202 struct tcp_options_received {
 203 /*      PAWS/RTTM data  */
 204         long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
 205         u32     ts_recent;      /* Time stamp to echo next              */
 206         u32     rcv_tsval;      /* Time stamp value                     */
 207         u32     rcv_tsecr;      /* Time stamp echo reply                */
 208         u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
 209                 tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
 210                 dsack : 1,      /* D-SACK is scheduled                  */
 211                 wscale_ok : 1,  /* Wscale seen on SYN packet            */
 212                 sack_ok : 4,    /* SACK seen on SYN packet              */
 213                 snd_wscale : 4, /* Window scaling received from sender  */
 214                 rcv_wscale : 4; /* Window scaling to send to receiver   */
 215 /*      SACKs data      */
 216         u8      eff_sacks;      /* Size of SACK array to send with next packet */
 217         u8      num_sacks;      /* Number of SACK blocks                */
 218         u16     user_mss;       /* mss requested by user in ioctl */
 219         u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
 220 };
 221
 222 struct tcp_request_sock {
 223         struct inet_request_sock        req;
 224 #ifdef CONFIG_TCP_MD5SIG
 225         /* Only used by TCP MD5 Signature so far. */
 226         struct tcp_request_sock_ops     *af_specific;
 227 #endif
 228         u32                             rcv_isn;
 229         u32                             snt_isn;
 230 };
 231
 232 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 233 {
 234         return (struct tcp_request_sock *)req;
 235 }
 236
 237 struct tcp_sock {
 238         /* inet_connection_sock has to be the first member of tcp_sock */
 239         struct inet_connection_sock     inet_conn;
 240         u16     tcp_header_len; /* Bytes of tcp header to send          */
 241         u16     xmit_size_goal; /* Goal for segmenting output packets   */
 242
 243 /*
 244  *      Header prediction flags
 245  *      0x5?10 << 16 + snd_wnd in net byte order
 246  */
 247         __be32  pred_flags;
 248
 249 /*
 250  *      RFC793 variables by their proper names. This means you can
 251  *      read the code and the spec side by side (and laugh ...)
 252  *      See RFC793 and RFC1122. The RFC writes these in capitals.
 253  */
 254         u32     rcv_nxt;        /* What we want to receive next         */
 255         u32     copied_seq;     /* Head of yet unread data              */
 256         u32     rcv_wup;        /* rcv_nxt on last window update sent   */
 257         u32     snd_nxt;        /* Next sequence we send                */
 258
 259         u32     snd_una;        /* First byte we want an ack for        */
 260         u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
 261         u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 262         u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 263
 264         /* Data for direct copy to user */
 265         struct {
 266                 struct sk_buff_head     prequeue;
 267                 struct task_struct      *task;
 268                 struct iovec            *iov;
 269                 int                     memory;
 270                 int                     len;
 271 #ifdef CONFIG_NET_DMA
 272                 /* members for async copy */
 273                 struct dma_chan         *dma_chan;
 274                 int                     wakeup;
 275                 struct dma_pinned_list  *pinned_list;
 276                 dma_cookie_t            dma_cookie;
 277 #endif
 278         } ucopy;
 279
 280         u32     snd_wl1;        /* Sequence for window update           */
 281         u32     snd_wnd;        /* The window we expect to receive      */
 282         u32     max_window;     /* Maximal window ever seen from peer   */
 283         u32     mss_cache;      /* Cached effective mss, not including SACKS */
 284
 285         u32     window_clamp;   /* Maximal window to advertise          */
 286         u32     rcv_ssthresh;   /* Current window clamp                 */
 287
 288         u32     frto_highmark;  /* snd_nxt when RTO occurred */
 289         u8      reordering;     /* Packet reordering metric.            */
 290         u8      frto_counter;   /* Number of new acks after RTO */
 291         u8      nonagle;        /* Disable Nagle algorithm?             */
 292         u8      keepalive_probes; /* num of allowed keep alive probes   */
 293
 294 /* RTT measurement */
 295         u32     srtt;           /* smoothed round trip time << 3        */
 296         u32     mdev;           /* medium deviation                     */
 297         u32     mdev_max;       /* maximal mdev for the last rtt period */
 298         u32     rttvar;         /* smoothed mdev_max                    */
 299         u32     rtt_seq;        /* sequence number to update rttvar     */
 300
 301         u32     packets_out;    /* Packets which are "in flight"        */
 302         u32     left_out;       /* Packets which leaved network */
 303         u32     retrans_out;    /* Retransmitted packets out            */
 304 /*
 305  *      Options received (usually on last packet, some only on SYN packets).
 306  */
 307         struct tcp_options_received rx_opt;
 308
 309 /*
 310  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 311  */
 312         u32     snd_ssthresh;   /* Slow start size threshold            */
 313         u32     snd_cwnd;       /* Sending congestion window            */
 314         u16     snd_cwnd_cnt;   /* Linear increase counter              */
 315         u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 316         u32     snd_cwnd_used;
 317         u32     snd_cwnd_stamp;
 318
 319         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 320
 321         u32     rcv_wnd;        /* Current receiver window              */
 322         u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 323         u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 324
 325 /*      SACKs data      */
 326         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 327         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 328
 329         struct tcp_sack_block_wire recv_sack_cache[4];
 330
 331         /* from STCP, retrans queue hinting */
 332         struct sk_buff* lost_skb_hint;
 333
 334         struct sk_buff *scoreboard_skb_hint;
 335         struct sk_buff *retransmit_skb_hint;
 336         struct sk_buff *forward_skb_hint;
 337         struct sk_buff *fastpath_skb_hint;
 338
 339         int     fastpath_cnt_hint;
 340         int     lost_cnt_hint;
 341         int     retransmit_cnt_hint;
 342         int     forward_cnt_hint;
 343
 344         u16     advmss;         /* Advertised MSS                       */
 345         u16     prior_ssthresh; /* ssthresh saved at recovery start     */
 346         u32     lost_out;       /* Lost packets                 */
 347         u32     sacked_out;     /* SACK'd packets                       */
 348         u32     fackets_out;    /* FACK'd packets                       */
 349         u32     high_seq;       /* snd_nxt at onset of congestion       */
 350
 351         u32     retrans_stamp;  /* Timestamp of the last retransmit,
 352                                  * also used in SYN-SENT to remember stamp of
 353                                  * the first SYN. */
 354         u32     undo_marker;    /* tracking retrans started here. */
 355         int     undo_retrans;   /* number of undoable retransmissions. */
 356         u32     urg_seq;        /* Seq of received urgent pointer */
 357         u16     urg_data;       /* Saved octet of OOB data and control flags */
 358         u8      urg_mode;       /* In urgent mode               */
 359         u8      ecn_flags;      /* ECN status bits.                     */
 360         u32     snd_up;         /* Urgent pointer               */
 361
 362         u32     total_retrans;  /* Total retransmits for entire connection */
 363         u32     bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
 364
 365         unsigned int            keepalive_time;   /* time before keep alive takes place */
 366         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 367         int                     linger2;
 368
 369         unsigned long last_synq_overflow;
 370
 371         u32     tso_deferred;
 372
 373 /* Receiver side RTT estimation */
 374         struct {
 375                 u32     rtt;
 376                 u32     seq;
 377                 u32     time;
 378         } rcv_rtt_est;
 379
 380 /* Receiver queue space */
 381         struct {
 382                 int     space;
 383                 u32     seq;
 384                 u32     time;
 385         } rcvq_space;
 386
 387 /* TCP-specific MTU probe information. */
 388         struct {
 389                 u32               probe_seq_start;
 390                 u32               probe_seq_end;
 391         } mtu_probe;
 392
 393 #ifdef CONFIG_TCP_MD5SIG
 394 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
 395         struct tcp_sock_af_ops  *af_specific;
 396
 397 /* TCP MD5 Signagure Option information */
 398         struct tcp_md5sig_info  *md5sig_info;
 399 #endif
 400 };
 401
 402 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 403 {
 404         return (struct tcp_sock *)sk;
 405 }
 406
 407 struct tcp_timewait_sock {
 408         struct inet_timewait_sock tw_sk;
 409         u32                       tw_rcv_nxt;
 410         u32                       tw_snd_nxt;
 411         u32                       tw_rcv_wnd;
 412         u32                       tw_ts_recent;
 413         long                      tw_ts_recent_stamp;
 414 #ifdef CONFIG_TCP_MD5SIG
 415         u16                       tw_md5_keylen;
 416         u8                        tw_md5_key[TCP_MD5SIG_MAXKEYLEN];
 417 #endif
 418 };
 419
 420 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 421 {
 422         return (struct tcp_timewait_sock *)sk;
 423 }
 424
 425 #endif
 426
 427 #endif  /* _LINUX_TCP_H */