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
  21#include <linux/skbuff.h>
  22#include <linux/dmaengine.h>
  23#include <net/sock.h>
  24#include <net/inet_connection_sock.h>
  25#include <net/inet_timewait_sock.h>
  26#include <uapi/linux/tcp.h>
  27
  28static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
  29{
  30        return (struct tcphdr *)skb_transport_header(skb);
  31}
  32
  33static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
  34{
  35        return tcp_hdr(skb)->doff * 4;
  36}
  37
  38static inline unsigned int tcp_optlen(const struct sk_buff *skb)
  39{
  40        return (tcp_hdr(skb)->doff - 5) * 4;
  41}
  42
  43/* TCP Fast Open */
  44#define TCP_FASTOPEN_COOKIE_MIN 4       /* Min Fast Open Cookie size in bytes */
  45#define TCP_FASTOPEN_COOKIE_MAX 16      /* Max Fast Open Cookie size in bytes */
  46#define TCP_FASTOPEN_COOKIE_SIZE 8      /* the size employed by this impl. */
  47
  48/* TCP Fast Open Cookie as stored in memory */
  49struct tcp_fastopen_cookie {
  50        s8      len;
  51        u8      val[TCP_FASTOPEN_COOKIE_MAX];
  52};
  53
  54/* This defines a selective acknowledgement block. */
  55struct tcp_sack_block_wire {
  56        __be32  start_seq;
  57        __be32  end_seq;
  58};
  59
  60struct tcp_sack_block {
  61        u32     start_seq;
  62        u32     end_seq;
  63};
  64
  65/*These are used to set the sack_ok field in struct tcp_options_received */
  66#define TCP_SACK_SEEN     (1 << 0)   /*1 = peer is SACK capable, */
  67#define TCP_FACK_ENABLED  (1 << 1)   /*1 = FACK is enabled locally*/
  68#define TCP_DSACK_SEEN    (1 << 2)   /*1 = DSACK was received from peer*/
  69
  70struct tcp_options_received {
  71/*      PAWS/RTTM data  */
  72        long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
  73        u32     ts_recent;      /* Time stamp to echo next              */
  74        u32     rcv_tsval;      /* Time stamp value                     */
  75        u32     rcv_tsecr;      /* Time stamp echo reply                */
  76        u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
  77                tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
  78                dsack : 1,      /* D-SACK is scheduled                  */
  79                wscale_ok : 1,  /* Wscale seen on SYN packet            */
  80                sack_ok : 4,    /* SACK seen on SYN packet              */
  81                snd_wscale : 4, /* Window scaling received from sender  */
  82                rcv_wscale : 4; /* Window scaling to send to receiver   */
  83        u8      cookie_plus:6,  /* bytes in authenticator/cookie option */
  84                cookie_out_never:1,
  85                cookie_in_always:1;
  86        u8      num_sacks;      /* Number of SACK blocks                */
  87        u16     user_mss;       /* mss requested by user in ioctl       */
  88        u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
  89};
  90
  91static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
  92{
  93        rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
  94        rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
  95        rx_opt->cookie_plus = 0;
  96}
  97
  98/* This is the max number of SACKS that we'll generate and process. It's safe
  99 * to increase this, although since:
 100 *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
 101 * only four options will fit in a standard TCP header */
 102#define TCP_NUM_SACKS 4
 103
 104struct tcp_cookie_values;
 105struct tcp_request_sock_ops;
 106
 107struct tcp_request_sock {
 108        struct inet_request_sock        req;
 109#ifdef CONFIG_TCP_MD5SIG
 110        /* Only used by TCP MD5 Signature so far. */
 111        const struct tcp_request_sock_ops *af_specific;
 112#endif
 113        struct sock                     *listener; /* needed for TFO */
 114        u32                             rcv_isn;
 115        u32                             snt_isn;
 116        u32                             snt_synack; /* synack sent time */
 117        u32                             rcv_nxt; /* the ack # by SYNACK. For
 118                                                  * FastOpen it's the seq#
 119                                                  * after data-in-SYN.
 120                                                  */
 121};
 122
 123static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 124{
 125        return (struct tcp_request_sock *)req;
 126}
 127
 128struct tcp_sock {
 129        /* inet_connection_sock has to be the first member of tcp_sock */
 130        struct inet_connection_sock     inet_conn;
 131        u16     tcp_header_len; /* Bytes of tcp header to send          */
 132        u16     xmit_size_goal_segs; /* Goal for segmenting output packets */
 133
 134/*
 135 *      Header prediction flags
 136 *      0x5?10 << 16 + snd_wnd in net byte order
 137 */
 138        __be32  pred_flags;
 139
 140/*
 141 *      RFC793 variables by their proper names. This means you can
 142 *      read the code and the spec side by side (and laugh ...)
 143 *      See RFC793 and RFC1122. The RFC writes these in capitals.
 144 */
 145        u32     rcv_nxt;        /* What we want to receive next         */
 146        u32     copied_seq;     /* Head of yet unread data              */
 147        u32     rcv_wup;        /* rcv_nxt on last window update sent   */
 148        u32     snd_nxt;        /* Next sequence we send                */
 149
 150        u32     snd_una;        /* First byte we want an ack for        */
 151        u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
 152        u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 153        u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 154
 155        struct list_head tsq_node; /* anchor in tsq_tasklet.head list */
 156        unsigned long   tsq_flags;
 157
 158        /* Data for direct copy to user */
 159        struct {
 160                struct sk_buff_head     prequeue;
 161                struct task_struct      *task;
 162                struct iovec            *iov;
 163                int                     memory;
 164                int                     len;
 165#ifdef CONFIG_NET_DMA
 166                /* members for async copy */
 167                struct dma_chan         *dma_chan;
 168                int                     wakeup;
 169                struct dma_pinned_list  *pinned_list;
 170                dma_cookie_t            dma_cookie;
 171#endif
 172        } ucopy;
 173
 174        u32     snd_wl1;        /* Sequence for window update           */
 175        u32     snd_wnd;        /* The window we expect to receive      */
 176        u32     max_window;     /* Maximal window ever seen from peer   */
 177        u32     mss_cache;      /* Cached effective mss, not including SACKS */
 178
 179        u32     window_clamp;   /* Maximal window to advertise          */
 180        u32     rcv_ssthresh;   /* Current window clamp                 */
 181
 182        u32     frto_highmark;  /* snd_nxt when RTO occurred */
 183        u16     advmss;         /* Advertised MSS                       */
 184        u8      frto_counter;   /* Number of new acks after RTO */
 185        u8      nonagle     : 4,/* Disable Nagle algorithm?             */
 186                thin_lto    : 1,/* Use linear timeouts for thin streams */
 187                thin_dupack : 1,/* Fast retransmit on first dupack      */
 188                repair      : 1,
 189                unused      : 1;
 190        u8      repair_queue;
 191        u8      do_early_retrans:1,/* Enable RFC5827 early-retransmit  */
 192                early_retrans_delayed:1, /* Delayed ER timer installed */
 193                syn_data:1,     /* SYN includes data */
 194                syn_fastopen:1, /* SYN includes Fast Open option */
 195                syn_data_acked:1;/* data in SYN is acked by SYN-ACK */
 196
 197/* RTT measurement */
 198        u32     srtt;           /* smoothed round trip time << 3        */
 199        u32     mdev;           /* medium deviation                     */
 200        u32     mdev_max;       /* maximal mdev for the last rtt period */
 201        u32     rttvar;         /* smoothed mdev_max                    */
 202        u32     rtt_seq;        /* sequence number to update rttvar     */
 203
 204        u32     packets_out;    /* Packets which are "in flight"        */
 205        u32     retrans_out;    /* Retransmitted packets out            */
 206
 207        u16     urg_data;       /* Saved octet of OOB data and control flags */
 208        u8      ecn_flags;      /* ECN status bits.                     */
 209        u8      reordering;     /* Packet reordering metric.            */
 210        u32     snd_up;         /* Urgent pointer               */
 211
 212        u8      keepalive_probes; /* num of allowed keep alive probes   */
 213/*
 214 *      Options received (usually on last packet, some only on SYN packets).
 215 */
 216        struct tcp_options_received rx_opt;
 217
 218/*
 219 *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 220 */
 221        u32     snd_ssthresh;   /* Slow start size threshold            */
 222        u32     snd_cwnd;       /* Sending congestion window            */
 223        u32     snd_cwnd_cnt;   /* Linear increase counter              */
 224        u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 225        u32     snd_cwnd_used;
 226        u32     snd_cwnd_stamp;
 227        u32     prior_cwnd;     /* Congestion window at start of Recovery. */
 228        u32     prr_delivered;  /* Number of newly delivered packets to
 229                                 * receiver in Recovery. */
 230        u32     prr_out;        /* Total number of pkts sent during Recovery. */
 231
 232        u32     rcv_wnd;        /* Current receiver window              */
 233        u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 234        u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 235        u32     lost_out;       /* Lost packets                 */
 236        u32     sacked_out;     /* SACK'd packets                       */
 237        u32     fackets_out;    /* FACK'd packets                       */
 238        u32     tso_deferred;
 239        u32     bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
 240
 241        /* from STCP, retrans queue hinting */
 242        struct sk_buff* lost_skb_hint;
 243        struct sk_buff *scoreboard_skb_hint;
 244        struct sk_buff *retransmit_skb_hint;
 245
 246        struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 247
 248        /* SACKs data, these 2 need to be together (see tcp_options_write) */
 249        struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 250        struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 251
 252        struct tcp_sack_block recv_sack_cache[4];
 253
 254        struct sk_buff *highest_sack;   /* skb just after the highest
 255                                         * skb with SACKed bit set
 256                                         * (validity guaranteed only if
 257                                         * sacked_out > 0)
 258                                         */
 259
 260        int     lost_cnt_hint;
 261        u32     retransmit_high;        /* L-bits may be on up to this seqno */
 262
 263        u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 264
 265        u32     prior_ssthresh; /* ssthresh saved at recovery start     */
 266        u32     high_seq;       /* snd_nxt at onset of congestion       */
 267
 268        u32     retrans_stamp;  /* Timestamp of the last retransmit,
 269                                 * also used in SYN-SENT to remember stamp of
 270                                 * the first SYN. */
 271        u32     undo_marker;    /* tracking retrans started here. */
 272        int     undo_retrans;   /* number of undoable retransmissions. */
 273        u32     total_retrans;  /* Total retransmits for entire connection */
 274
 275        u32     urg_seq;        /* Seq of received urgent pointer */
 276        unsigned int            keepalive_time;   /* time before keep alive takes place */
 277        unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 278
 279        int                     linger2;
 280
 281/* Receiver side RTT estimation */
 282        struct {
 283                u32     rtt;
 284                u32     seq;
 285                u32     time;
 286        } rcv_rtt_est;
 287
 288/* Receiver queue space */
 289        struct {
 290                int     space;
 291                u32     seq;
 292                u32     time;
 293        } rcvq_space;
 294
 295/* TCP-specific MTU probe information. */
 296        struct {
 297                u32               probe_seq_start;
 298                u32               probe_seq_end;
 299        } mtu_probe;
 300        u32     mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
 301                           * while socket was owned by user.
 302                           */
 303
 304#ifdef CONFIG_TCP_MD5SIG
 305/* TCP AF-Specific parts; only used by MD5 Signature support so far */
 306        const struct tcp_sock_af_ops    *af_specific;
 307
 308/* TCP MD5 Signature Option information */
 309        struct tcp_md5sig_info  __rcu *md5sig_info;
 310#endif
 311
 312        /* When the cookie options are generated and exchanged, then this
 313         * object holds a reference to them (cookie_values->kref).  Also
 314         * contains related tcp_cookie_transactions fields.
 315         */
 316        struct tcp_cookie_values  *cookie_values;
 317
 318/* TCP fastopen related information */
 319        struct tcp_fastopen_request *fastopen_req;
 320        /* fastopen_rsk points to request_sock that resulted in this big
 321         * socket. Used to retransmit SYNACKs etc.
 322         */
 323        struct request_sock *fastopen_rsk;
 324};
 325
 326enum tsq_flags {
 327        TSQ_THROTTLED,
 328        TSQ_QUEUED,
 329        TCP_TSQ_DEFERRED,          /* tcp_tasklet_func() found socket was owned */
 330        TCP_WRITE_TIMER_DEFERRED,  /* tcp_write_timer() found socket was owned */
 331        TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */
 332        TCP_MTU_REDUCED_DEFERRED,  /* tcp_v{4|6}_err() could not call
 333                                    * tcp_v{4|6}_mtu_reduced()
 334                                    */
 335};
 336
 337static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 338{
 339        return (struct tcp_sock *)sk;
 340}
 341
 342struct tcp_timewait_sock {
 343        struct inet_timewait_sock tw_sk;
 344        u32                       tw_rcv_nxt;
 345        u32                       tw_snd_nxt;
 346        u32                       tw_rcv_wnd;
 347        u32                       tw_ts_recent;
 348        long                      tw_ts_recent_stamp;
 349#ifdef CONFIG_TCP_MD5SIG
 350        struct tcp_md5sig_key     *tw_md5_key;
 351#endif
 352        /* Few sockets in timewait have cookies; in that case, then this
 353         * object holds a reference to them (tw_cookie_values->kref).
 354         */
 355        struct tcp_cookie_values  *tw_cookie_values;
 356};
 357
 358static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 359{
 360        return (struct tcp_timewait_sock *)sk;
 361}
 362
 363static inline bool tcp_passive_fastopen(const struct sock *sk)
 364{
 365        return (sk->sk_state == TCP_SYN_RECV &&
 366                tcp_sk(sk)->fastopen_rsk != NULL);
 367}
 368
 369static inline bool fastopen_cookie_present(struct tcp_fastopen_cookie *foc)
 370{
 371        return foc->len != -1;
 372}
 373
 374extern void tcp_sock_destruct(struct sock *sk);
 375
 376static inline int fastopen_init_queue(struct sock *sk, int backlog)
 377{
 378        struct request_sock_queue *queue =
 379            &inet_csk(sk)->icsk_accept_queue;
 380
 381        if (queue->fastopenq == NULL) {
 382                queue->fastopenq = kzalloc(
 383                    sizeof(struct fastopen_queue),
 384                    sk->sk_allocation);
 385                if (queue->fastopenq == NULL)
 386                        return -ENOMEM;
 387
 388                sk->sk_destruct = tcp_sock_destruct;
 389                spin_lock_init(&queue->fastopenq->lock);
 390        }
 391        queue->fastopenq->max_qlen = backlog;
 392        return 0;
 393}
 394
 395#endif  /* _LINUX_TCP_H */
 396