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_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 /* This defines a selective acknowledgement block. */
 247 struct tcp_sack_block_wire {
 248         __be32  start_seq;
 249         __be32  end_seq;
 250 };
 251
 252 struct tcp_sack_block {
 253         u32     start_seq;
 254         u32     end_seq;
 255 };
 256
 257 /*These are used to set the sack_ok field in struct tcp_options_received */
 258 #define TCP_SACK_SEEN     (1 << 0)   /*1 = peer is SACK capable, */
 259 #define TCP_FACK_ENABLED  (1 << 1)   /*1 = FACK is enabled locally*/
 260 #define TCP_DSACK_SEEN    (1 << 2)   /*1 = DSACK was received from peer*/
 261
 262 struct tcp_options_received {
 263 /*      PAWS/RTTM data  */
 264         long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
 265         u32     ts_recent;      /* Time stamp to echo next              */
 266         u32     rcv_tsval;      /* Time stamp value                     */
 267         u32     rcv_tsecr;      /* Time stamp echo reply                */
 268         u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
 269                 tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
 270                 dsack : 1,      /* D-SACK is scheduled                  */
 271                 wscale_ok : 1,  /* Wscale seen on SYN packet            */
 272                 sack_ok : 4,    /* SACK seen on SYN packet              */
 273                 snd_wscale : 4, /* Window scaling received from sender  */
 274                 rcv_wscale : 4; /* Window scaling to send to receiver   */
 275         u8      cookie_plus:6,  /* bytes in authenticator/cookie option */
 276                 cookie_out_never:1,
 277                 cookie_in_always:1;
 278         u8      num_sacks;      /* Number of SACK blocks                */
 279         u16     user_mss;       /* mss requested by user in ioctl       */
 280         u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
 281 };
 282
 283 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
 284 {
 285         rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
 286         rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
 287         rx_opt->cookie_plus = 0;
 288 }
 289
 290 /* This is the max number of SACKS that we'll generate and process. It's safe
 291  * to increase this, although since:
 292  *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
 293  * only four options will fit in a standard TCP header */
 294 #define TCP_NUM_SACKS 4
 295
 296 struct tcp_cookie_values;
 297 struct tcp_request_sock_ops;
 298
 299 struct tcp_request_sock {
 300         struct inet_request_sock        req;
 301 #ifdef CONFIG_TCP_MD5SIG
 302         /* Only used by TCP MD5 Signature so far. */
 303         const struct tcp_request_sock_ops *af_specific;
 304 #endif
 305         u32                             rcv_isn;
 306         u32                             snt_isn;
 307         u32                             snt_synack; /* synack sent time */
 308 };
 309
 310 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 311 {
 312         return (struct tcp_request_sock *)req;
 313 }
 314
 315 struct tcp_sock {
 316         /* inet_connection_sock has to be the first member of tcp_sock */
 317         struct inet_connection_sock     inet_conn;
 318         u16     tcp_header_len; /* Bytes of tcp header to send          */
 319         u16     xmit_size_goal_segs; /* Goal for segmenting output packets */
 320
 321 /*
 322  *      Header prediction flags
 323  *      0x5?10 << 16 + snd_wnd in net byte order
 324  */
 325         __be32  pred_flags;
 326
 327 /*
 328  *      RFC793 variables by their proper names. This means you can
 329  *      read the code and the spec side by side (and laugh ...)
 330  *      See RFC793 and RFC1122. The RFC writes these in capitals.
 331  */
 332         u32     rcv_nxt;        /* What we want to receive next         */
 333         u32     copied_seq;     /* Head of yet unread data              */
 334         u32     rcv_wup;        /* rcv_nxt on last window update sent   */
 335         u32     snd_nxt;        /* Next sequence we send                */
 336
 337         u32     snd_una;        /* First byte we want an ack for        */
 338         u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
 339         u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 340         u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 341
 342         /* Data for direct copy to user */
 343         struct {
 344                 struct sk_buff_head     prequeue;
 345                 struct task_struct      *task;
 346                 struct iovec            *iov;
 347                 int                     memory;
 348                 int                     len;
 349 #ifdef CONFIG_NET_DMA
 350                 /* members for async copy */
 351                 struct dma_chan         *dma_chan;
 352                 int                     wakeup;
 353                 struct dma_pinned_list  *pinned_list;
 354                 dma_cookie_t            dma_cookie;
 355 #endif
 356         } ucopy;
 357
 358         u32     snd_wl1;        /* Sequence for window update           */
 359         u32     snd_wnd;        /* The window we expect to receive      */
 360         u32     max_window;     /* Maximal window ever seen from peer   */
 361         u32     mss_cache;      /* Cached effective mss, not including SACKS */
 362
 363         u32     window_clamp;   /* Maximal window to advertise          */
 364         u32     rcv_ssthresh;   /* Current window clamp                 */
 365
 366         u32     frto_highmark;  /* snd_nxt when RTO occurred */
 367         u16     advmss;         /* Advertised MSS                       */
 368         u8      frto_counter;   /* Number of new acks after RTO */
 369         u8      nonagle     : 4,/* Disable Nagle algorithm?             */
 370                 thin_lto    : 1,/* Use linear timeouts for thin streams */
 371                 thin_dupack : 1,/* Fast retransmit on first dupack      */
 372                 repair      : 1,
 373                 unused      : 1;
 374         u8      repair_queue;
 375         u8      do_early_retrans:1,/* Enable RFC5827 early-retransmit  */
 376                 early_retrans_delayed:1; /* Delayed ER timer installed */
 377
 378 /* RTT measurement */
 379         u32     srtt;           /* smoothed round trip time << 3        */
 380         u32     mdev;           /* medium deviation                     */
 381         u32     mdev_max;       /* maximal mdev for the last rtt period */
 382         u32     rttvar;         /* smoothed mdev_max                    */
 383         u32     rtt_seq;        /* sequence number to update rttvar     */
 384
 385         u32     packets_out;    /* Packets which are "in flight"        */
 386         u32     retrans_out;    /* Retransmitted packets out            */
 387
 388         u16     urg_data;       /* Saved octet of OOB data and control flags */
 389         u8      ecn_flags;      /* ECN status bits.                     */
 390         u8      reordering;     /* Packet reordering metric.            */
 391         u32     snd_up;         /* Urgent pointer               */
 392
 393         u8      keepalive_probes; /* num of allowed keep alive probes   */
 394 /*
 395  *      Options received (usually on last packet, some only on SYN packets).
 396  */
 397         struct tcp_options_received rx_opt;
 398
 399 /*
 400  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 401  */
 402         u32     snd_ssthresh;   /* Slow start size threshold            */
 403         u32     snd_cwnd;       /* Sending congestion window            */
 404         u32     snd_cwnd_cnt;   /* Linear increase counter              */
 405         u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 406         u32     snd_cwnd_used;
 407         u32     snd_cwnd_stamp;
 408         u32     prior_cwnd;     /* Congestion window at start of Recovery. */
 409         u32     prr_delivered;  /* Number of newly delivered packets to
 410                                  * receiver in Recovery. */
 411         u32     prr_out;        /* Total number of pkts sent during Recovery. */
 412
 413         u32     rcv_wnd;        /* Current receiver window              */
 414         u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 415         u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 416         u32     lost_out;       /* Lost packets                 */
 417         u32     sacked_out;     /* SACK'd packets                       */
 418         u32     fackets_out;    /* FACK'd packets                       */
 419         u32     tso_deferred;
 420         u32     bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
 421
 422         /* from STCP, retrans queue hinting */
 423         struct sk_buff* lost_skb_hint;
 424         struct sk_buff *scoreboard_skb_hint;
 425         struct sk_buff *retransmit_skb_hint;
 426
 427         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 428
 429         /* SACKs data, these 2 need to be together (see tcp_options_write) */
 430         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 431         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 432
 433         struct tcp_sack_block recv_sack_cache[4];
 434
 435         struct sk_buff *highest_sack;   /* skb just after the highest
 436                                          * skb with SACKed bit set
 437                                          * (validity guaranteed only if
 438                                          * sacked_out > 0)
 439                                          */
 440
 441         int     lost_cnt_hint;
 442         u32     retransmit_high;        /* L-bits may be on up to this seqno */
 443
 444         u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 445
 446         u32     prior_ssthresh; /* ssthresh saved at recovery start     */
 447         u32     high_seq;       /* snd_nxt at onset of congestion       */
 448
 449         u32     retrans_stamp;  /* Timestamp of the last retransmit,
 450                                  * also used in SYN-SENT to remember stamp of
 451                                  * the first SYN. */
 452         u32     undo_marker;    /* tracking retrans started here. */
 453         int     undo_retrans;   /* number of undoable retransmissions. */
 454         u32     total_retrans;  /* Total retransmits for entire connection */
 455
 456         u32     urg_seq;        /* Seq of received urgent pointer */
 457         unsigned int            keepalive_time;   /* time before keep alive takes place */
 458         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 459
 460         int                     linger2;
 461
 462 /* Receiver side RTT estimation */
 463         struct {
 464                 u32     rtt;
 465                 u32     seq;
 466                 u32     time;
 467         } rcv_rtt_est;
 468
 469 /* Receiver queue space */
 470         struct {
 471                 int     space;
 472                 u32     seq;
 473                 u32     time;
 474         } rcvq_space;
 475
 476 /* TCP-specific MTU probe information. */
 477         struct {
 478                 u32               probe_seq_start;
 479                 u32               probe_seq_end;
 480         } mtu_probe;
 481
 482 #ifdef CONFIG_TCP_MD5SIG
 483 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
 484         const struct tcp_sock_af_ops    *af_specific;
 485
 486 /* TCP MD5 Signature Option information */
 487         struct tcp_md5sig_info  __rcu *md5sig_info;
 488 #endif
 489
 490         /* When the cookie options are generated and exchanged, then this
 491          * object holds a reference to them (cookie_values->kref).  Also
 492          * contains related tcp_cookie_transactions fields.
 493          */
 494         struct tcp_cookie_values  *cookie_values;
 495 };
 496
 497 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 498 {
 499         return (struct tcp_sock *)sk;
 500 }
 501
 502 struct tcp_timewait_sock {
 503         struct inet_timewait_sock tw_sk;
 504         u32                       tw_rcv_nxt;
 505         u32                       tw_snd_nxt;
 506         u32                       tw_rcv_wnd;
 507         u32                       tw_ts_recent;
 508         long                      tw_ts_recent_stamp;
 509 #ifdef CONFIG_TCP_MD5SIG
 510         struct tcp_md5sig_key   *tw_md5_key;
 511 #endif
 512         /* Few sockets in timewait have cookies; in that case, then this
 513          * object holds a reference to them (tw_cookie_values->kref).
 514          */
 515         struct tcp_cookie_values  *tw_cookie_values;
 516 };
 517
 518 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 519 {
 520         return (struct tcp_timewait_sock *)sk;
 521 }
 522
 523 #endif  /* __KERNEL__ */
 524
 525 #endif  /* _LINUX_TCP_H */