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
  28 static inline struct tcphdr *tcp_hdr(const struct sk_buff *skb)
  29 {
  30         return (struct tcphdr *)skb_transport_header(skb);
  31 }
  32
  33 static inline unsigned int tcp_hdrlen(const struct sk_buff *skb)
  34 {
  35         return tcp_hdr(skb)->doff * 4;
  36 }
  37
  38 static 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 */
  49 struct tcp_fastopen_cookie {
  50         s8      len;
  51         u8      val[TCP_FASTOPEN_COOKIE_MAX];
  52 };
  53
  54 /* This defines a selective acknowledgement block. */
  55 struct tcp_sack_block_wire {
  56         __be32  start_seq;
  57         __be32  end_seq;
  58 };
  59
  60 struct 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
  70 struct 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
  91 static 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
 104 struct tcp_cookie_values;
 105 struct tcp_request_sock_ops;
 106
 107 struct 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
 123 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 124 {
 125         return (struct tcp_request_sock *)req;
 126 }
 127
 128 struct 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
 196 /* RTT measurement */
 197         u32     srtt;           /* smoothed round trip time << 3        */
 198         u32     mdev;           /* medium deviation                     */
 199         u32     mdev_max;       /* maximal mdev for the last rtt period */
 200         u32     rttvar;         /* smoothed mdev_max                    */
 201         u32     rtt_seq;        /* sequence number to update rttvar     */
 202
 203         u32     packets_out;    /* Packets which are "in flight"        */
 204         u32     retrans_out;    /* Retransmitted packets out            */
 205
 206         u16     urg_data;       /* Saved octet of OOB data and control flags */
 207         u8      ecn_flags;      /* ECN status bits.                     */
 208         u8      reordering;     /* Packet reordering metric.            */
 209         u32     snd_up;         /* Urgent pointer               */
 210
 211         u8      keepalive_probes; /* num of allowed keep alive probes   */
 212 /*
 213  *      Options received (usually on last packet, some only on SYN packets).
 214  */
 215         struct tcp_options_received rx_opt;
 216
 217 /*
 218  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 219  */
 220         u32     snd_ssthresh;   /* Slow start size threshold            */
 221         u32     snd_cwnd;       /* Sending congestion window            */
 222         u32     snd_cwnd_cnt;   /* Linear increase counter              */
 223         u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 224         u32     snd_cwnd_used;
 225         u32     snd_cwnd_stamp;
 226         u32     prior_cwnd;     /* Congestion window at start of Recovery. */
 227         u32     prr_delivered;  /* Number of newly delivered packets to
 228                                  * receiver in Recovery. */
 229         u32     prr_out;        /* Total number of pkts sent during Recovery. */
 230
 231         u32     rcv_wnd;        /* Current receiver window              */
 232         u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 233         u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 234         u32     lost_out;       /* Lost packets                 */
 235         u32     sacked_out;     /* SACK'd packets                       */
 236         u32     fackets_out;    /* FACK'd packets                       */
 237         u32     tso_deferred;
 238         u32     bytes_acked;    /* Appropriate Byte Counting - RFC3465 */
 239
 240         /* from STCP, retrans queue hinting */
 241         struct sk_buff* lost_skb_hint;
 242         struct sk_buff *scoreboard_skb_hint;
 243         struct sk_buff *retransmit_skb_hint;
 244
 245         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 246
 247         /* SACKs data, these 2 need to be together (see tcp_options_write) */
 248         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 249         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 250
 251         struct tcp_sack_block recv_sack_cache[4];
 252
 253         struct sk_buff *highest_sack;   /* skb just after the highest
 254                                          * skb with SACKed bit set
 255                                          * (validity guaranteed only if
 256                                          * sacked_out > 0)
 257                                          */
 258
 259         int     lost_cnt_hint;
 260         u32     retransmit_high;        /* L-bits may be on up to this seqno */
 261
 262         u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 263
 264         u32     prior_ssthresh; /* ssthresh saved at recovery start     */
 265         u32     high_seq;       /* snd_nxt at onset of congestion       */
 266
 267         u32     retrans_stamp;  /* Timestamp of the last retransmit,
 268                                  * also used in SYN-SENT to remember stamp of
 269                                  * the first SYN. */
 270         u32     undo_marker;    /* tracking retrans started here. */
 271         int     undo_retrans;   /* number of undoable retransmissions. */
 272         u32     total_retrans;  /* Total retransmits for entire connection */
 273
 274         u32     urg_seq;        /* Seq of received urgent pointer */
 275         unsigned int            keepalive_time;   /* time before keep alive takes place */
 276         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 277
 278         int                     linger2;
 279
 280 /* Receiver side RTT estimation */
 281         struct {
 282                 u32     rtt;
 283                 u32     seq;
 284                 u32     time;
 285         } rcv_rtt_est;
 286
 287 /* Receiver queue space */
 288         struct {
 289                 int     space;
 290                 u32     seq;
 291                 u32     time;
 292         } rcvq_space;
 293
 294 /* TCP-specific MTU probe information. */
 295         struct {
 296                 u32               probe_seq_start;
 297                 u32               probe_seq_end;
 298         } mtu_probe;
 299         u32     mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
 300                            * while socket was owned by user.
 301                            */
 302
 303 #ifdef CONFIG_TCP_MD5SIG
 304 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
 305         const struct tcp_sock_af_ops    *af_specific;
 306
 307 /* TCP MD5 Signature Option information */
 308         struct tcp_md5sig_info  __rcu *md5sig_info;
 309 #endif
 310
 311         /* When the cookie options are generated and exchanged, then this
 312          * object holds a reference to them (cookie_values->kref).  Also
 313          * contains related tcp_cookie_transactions fields.
 314          */
 315         struct tcp_cookie_values  *cookie_values;
 316
 317 /* TCP fastopen related information */
 318         struct tcp_fastopen_request *fastopen_req;
 319         /* fastopen_rsk points to request_sock that resulted in this big
 320          * socket. Used to retransmit SYNACKs etc.
 321          */
 322         struct request_sock *fastopen_rsk;
 323 };
 324
 325 enum tsq_flags {
 326         TSQ_THROTTLED,
 327         TSQ_QUEUED,
 328         TCP_TSQ_DEFERRED,          /* tcp_tasklet_func() found socket was owned */
 329         TCP_WRITE_TIMER_DEFERRED,  /* tcp_write_timer() found socket was owned */
 330         TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */
 331         TCP_MTU_REDUCED_DEFERRED,  /* tcp_v{4|6}_err() could not call
 332                                     * tcp_v{4|6}_mtu_reduced()
 333                                     */
 334 };
 335
 336 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 337 {
 338         return (struct tcp_sock *)sk;
 339 }
 340
 341 struct tcp_timewait_sock {
 342         struct inet_timewait_sock tw_sk;
 343         u32                       tw_rcv_nxt;
 344         u32                       tw_snd_nxt;
 345         u32                       tw_rcv_wnd;
 346         u32                       tw_ts_recent;
 347         long                      tw_ts_recent_stamp;
 348 #ifdef CONFIG_TCP_MD5SIG
 349         struct tcp_md5sig_key     *tw_md5_key;
 350 #endif
 351         /* Few sockets in timewait have cookies; in that case, then this
 352          * object holds a reference to them (tw_cookie_values->kref).
 353          */
 354         struct tcp_cookie_values  *tw_cookie_values;
 355 };
 356
 357 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 358 {
 359         return (struct tcp_timewait_sock *)sk;
 360 }
 361
 362 static inline bool tcp_passive_fastopen(const struct sock *sk)
 363 {
 364         return (sk->sk_state == TCP_SYN_RECV &&
 365                 tcp_sk(sk)->fastopen_rsk != NULL);
 366 }
 367
 368 static inline bool fastopen_cookie_present(struct tcp_fastopen_cookie *foc)
 369 {
 370         return foc->len != -1;
 371 }
 372
 373 extern void tcp_sock_destruct(struct sock *sk);
 374
 375 static inline int fastopen_init_queue(struct sock *sk, int backlog)
 376 {
 377         struct request_sock_queue *queue =
 378             &inet_csk(sk)->icsk_accept_queue;
 379
 380         if (queue->fastopenq == NULL) {
 381                 queue->fastopenq = kzalloc(
 382                     sizeof(struct fastopen_queue),
 383                     sk->sk_allocation);
 384                 if (queue->fastopenq == NULL)
 385                         return -ENOMEM;
 386
 387                 sk->sk_destruct = tcp_sock_destruct;
 388                 spin_lock_init(&queue->fastopenq->lock);
 389         }
 390         queue->fastopenq->max_qlen = backlog;
 391         return 0;
 392 }
 393
 394 #endif  /* _LINUX_TCP_H */