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 struct tcphdr *inner_tcp_hdr(const struct sk_buff *skb)
  39 {
  40         return (struct tcphdr *)skb_inner_transport_header(skb);
  41 }
  42
  43 static inline unsigned int inner_tcp_hdrlen(const struct sk_buff *skb)
  44 {
  45         return inner_tcp_hdr(skb)->doff * 4;
  46 }
  47
  48 static inline unsigned int tcp_optlen(const struct sk_buff *skb)
  49 {
  50         return (tcp_hdr(skb)->doff - 5) * 4;
  51 }
  52
  53 /* TCP Fast Open */
  54 #define TCP_FASTOPEN_COOKIE_MIN 4       /* Min Fast Open Cookie size in bytes */
  55 #define TCP_FASTOPEN_COOKIE_MAX 16      /* Max Fast Open Cookie size in bytes */
  56 #define TCP_FASTOPEN_COOKIE_SIZE 8      /* the size employed by this impl. */
  57
  58 /* TCP Fast Open Cookie as stored in memory */
  59 struct tcp_fastopen_cookie {
  60         s8      len;
  61         u8      val[TCP_FASTOPEN_COOKIE_MAX];
  62 };
  63
  64 /* This defines a selective acknowledgement block. */
  65 struct tcp_sack_block_wire {
  66         __be32  start_seq;
  67         __be32  end_seq;
  68 };
  69
  70 struct tcp_sack_block {
  71         u32     start_seq;
  72         u32     end_seq;
  73 };
  74
  75 /*These are used to set the sack_ok field in struct tcp_options_received */
  76 #define TCP_SACK_SEEN     (1 << 0)   /*1 = peer is SACK capable, */
  77 #define TCP_FACK_ENABLED  (1 << 1)   /*1 = FACK is enabled locally*/
  78 #define TCP_DSACK_SEEN    (1 << 2)   /*1 = DSACK was received from peer*/
  79
  80 struct tcp_options_received {
  81 /*      PAWS/RTTM data  */
  82         long    ts_recent_stamp;/* Time we stored ts_recent (for aging) */
  83         u32     ts_recent;      /* Time stamp to echo next              */
  84         u32     rcv_tsval;      /* Time stamp value                     */
  85         u32     rcv_tsecr;      /* Time stamp echo reply                */
  86         u16     saw_tstamp : 1, /* Saw TIMESTAMP on last packet         */
  87                 tstamp_ok : 1,  /* TIMESTAMP seen on SYN packet         */
  88                 dsack : 1,      /* D-SACK is scheduled                  */
  89                 wscale_ok : 1,  /* Wscale seen on SYN packet            */
  90                 sack_ok : 4,    /* SACK seen on SYN packet              */
  91                 snd_wscale : 4, /* Window scaling received from sender  */
  92                 rcv_wscale : 4; /* Window scaling to send to receiver   */
  93         u8      cookie_plus:6,  /* bytes in authenticator/cookie option */
  94                 cookie_out_never:1,
  95                 cookie_in_always:1;
  96         u8      num_sacks;      /* Number of SACK blocks                */
  97         u16     user_mss;       /* mss requested by user in ioctl       */
  98         u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
  99 };
 100
 101 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
 102 {
 103         rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
 104         rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
 105         rx_opt->cookie_plus = 0;
 106 }
 107
 108 /* This is the max number of SACKS that we'll generate and process. It's safe
 109  * to increase this, although since:
 110  *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
 111  * only four options will fit in a standard TCP header */
 112 #define TCP_NUM_SACKS 4
 113
 114 struct tcp_cookie_values;
 115 struct tcp_request_sock_ops;
 116
 117 struct tcp_request_sock {
 118         struct inet_request_sock        req;
 119 #ifdef CONFIG_TCP_MD5SIG
 120         /* Only used by TCP MD5 Signature so far. */
 121         const struct tcp_request_sock_ops *af_specific;
 122 #endif
 123         struct sock                     *listener; /* needed for TFO */
 124         u32                             rcv_isn;
 125         u32                             snt_isn;
 126         u32                             snt_synack; /* synack sent time */
 127         u32                             rcv_nxt; /* the ack # by SYNACK. For
 128                                                   * FastOpen it's the seq#
 129                                                   * after data-in-SYN.
 130                                                   */
 131 };
 132
 133 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 134 {
 135         return (struct tcp_request_sock *)req;
 136 }
 137
 138 struct tcp_sock {
 139         /* inet_connection_sock has to be the first member of tcp_sock */
 140         struct inet_connection_sock     inet_conn;
 141         u16     tcp_header_len; /* Bytes of tcp header to send          */
 142         u16     xmit_size_goal_segs; /* Goal for segmenting output packets */
 143
 144 /*
 145  *      Header prediction flags
 146  *      0x5?10 << 16 + snd_wnd in net byte order
 147  */
 148         __be32  pred_flags;
 149
 150 /*
 151  *      RFC793 variables by their proper names. This means you can
 152  *      read the code and the spec side by side (and laugh ...)
 153  *      See RFC793 and RFC1122. The RFC writes these in capitals.
 154  */
 155         u32     rcv_nxt;        /* What we want to receive next         */
 156         u32     copied_seq;     /* Head of yet unread data              */
 157         u32     rcv_wup;        /* rcv_nxt on last window update sent   */
 158         u32     snd_nxt;        /* Next sequence we send                */
 159
 160         u32     snd_una;        /* First byte we want an ack for        */
 161         u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
 162         u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 163         u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 164
 165         struct list_head tsq_node; /* anchor in tsq_tasklet.head list */
 166         unsigned long   tsq_flags;
 167
 168         /* Data for direct copy to user */
 169         struct {
 170                 struct sk_buff_head     prequeue;
 171                 struct task_struct      *task;
 172                 struct iovec            *iov;
 173                 int                     memory;
 174                 int                     len;
 175 #ifdef CONFIG_NET_DMA
 176                 /* members for async copy */
 177                 struct dma_chan         *dma_chan;
 178                 int                     wakeup;
 179                 struct dma_pinned_list  *pinned_list;
 180                 dma_cookie_t            dma_cookie;
 181 #endif
 182         } ucopy;
 183
 184         u32     snd_wl1;        /* Sequence for window update           */
 185         u32     snd_wnd;        /* The window we expect to receive      */
 186         u32     max_window;     /* Maximal window ever seen from peer   */
 187         u32     mss_cache;      /* Cached effective mss, not including SACKS */
 188
 189         u32     window_clamp;   /* Maximal window to advertise          */
 190         u32     rcv_ssthresh;   /* Current window clamp                 */
 191
 192         u32     frto_highmark;  /* snd_nxt when RTO occurred */
 193         u16     advmss;         /* Advertised MSS                       */
 194         u8      frto_counter;   /* Number of new acks after RTO */
 195         u8      nonagle     : 4,/* Disable Nagle algorithm?             */
 196                 thin_lto    : 1,/* Use linear timeouts for thin streams */
 197                 thin_dupack : 1,/* Fast retransmit on first dupack      */
 198                 repair      : 1,
 199                 unused      : 1;
 200         u8      repair_queue;
 201         u8      do_early_retrans:1,/* Enable RFC5827 early-retransmit  */
 202                 early_retrans_delayed:1, /* Delayed ER timer installed */
 203                 syn_data:1,     /* SYN includes data */
 204                 syn_fastopen:1, /* SYN includes Fast Open option */
 205                 syn_data_acked:1;/* data in SYN is acked by SYN-ACK */
 206
 207 /* RTT measurement */
 208         u32     srtt;           /* smoothed round trip time << 3        */
 209         u32     mdev;           /* medium deviation                     */
 210         u32     mdev_max;       /* maximal mdev for the last rtt period */
 211         u32     rttvar;         /* smoothed mdev_max                    */
 212         u32     rtt_seq;        /* sequence number to update rttvar     */
 213
 214         u32     packets_out;    /* Packets which are "in flight"        */
 215         u32     retrans_out;    /* Retransmitted packets out            */
 216
 217         u16     urg_data;       /* Saved octet of OOB data and control flags */
 218         u8      ecn_flags;      /* ECN status bits.                     */
 219         u8      reordering;     /* Packet reordering metric.            */
 220         u32     snd_up;         /* Urgent pointer               */
 221
 222         u8      keepalive_probes; /* num of allowed keep alive probes   */
 223 /*
 224  *      Options received (usually on last packet, some only on SYN packets).
 225  */
 226         struct tcp_options_received rx_opt;
 227
 228 /*
 229  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 230  */
 231         u32     snd_ssthresh;   /* Slow start size threshold            */
 232         u32     snd_cwnd;       /* Sending congestion window            */
 233         u32     snd_cwnd_cnt;   /* Linear increase counter              */
 234         u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 235         u32     snd_cwnd_used;
 236         u32     snd_cwnd_stamp;
 237         u32     prior_cwnd;     /* Congestion window at start of Recovery. */
 238         u32     prr_delivered;  /* Number of newly delivered packets to
 239                                  * receiver in Recovery. */
 240         u32     prr_out;        /* Total number of pkts sent during Recovery. */
 241
 242         u32     rcv_wnd;        /* Current receiver window              */
 243         u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 244         u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 245         u32     lost_out;       /* Lost packets                 */
 246         u32     sacked_out;     /* SACK'd packets                       */
 247         u32     fackets_out;    /* FACK'd packets                       */
 248         u32     tso_deferred;
 249
 250         /* from STCP, retrans queue hinting */
 251         struct sk_buff* lost_skb_hint;
 252         struct sk_buff *scoreboard_skb_hint;
 253         struct sk_buff *retransmit_skb_hint;
 254
 255         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 256
 257         /* SACKs data, these 2 need to be together (see tcp_options_write) */
 258         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 259         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 260
 261         struct tcp_sack_block recv_sack_cache[4];
 262
 263         struct sk_buff *highest_sack;   /* skb just after the highest
 264                                          * skb with SACKed bit set
 265                                          * (validity guaranteed only if
 266                                          * sacked_out > 0)
 267                                          */
 268
 269         int     lost_cnt_hint;
 270         u32     retransmit_high;        /* L-bits may be on up to this seqno */
 271
 272         u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 273
 274         u32     prior_ssthresh; /* ssthresh saved at recovery start     */
 275         u32     high_seq;       /* snd_nxt at onset of congestion       */
 276
 277         u32     retrans_stamp;  /* Timestamp of the last retransmit,
 278                                  * also used in SYN-SENT to remember stamp of
 279                                  * the first SYN. */
 280         u32     undo_marker;    /* tracking retrans started here. */
 281         int     undo_retrans;   /* number of undoable retransmissions. */
 282         u32     total_retrans;  /* Total retransmits for entire connection */
 283
 284         u32     urg_seq;        /* Seq of received urgent pointer */
 285         unsigned int            keepalive_time;   /* time before keep alive takes place */
 286         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 287
 288         int                     linger2;
 289
 290 /* Receiver side RTT estimation */
 291         struct {
 292                 u32     rtt;
 293                 u32     seq;
 294                 u32     time;
 295         } rcv_rtt_est;
 296
 297 /* Receiver queue space */
 298         struct {
 299                 int     space;
 300                 u32     seq;
 301                 u32     time;
 302         } rcvq_space;
 303
 304 /* TCP-specific MTU probe information. */
 305         struct {
 306                 u32               probe_seq_start;
 307                 u32               probe_seq_end;
 308         } mtu_probe;
 309         u32     mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
 310                            * while socket was owned by user.
 311                            */
 312
 313 #ifdef CONFIG_TCP_MD5SIG
 314 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
 315         const struct tcp_sock_af_ops    *af_specific;
 316
 317 /* TCP MD5 Signature Option information */
 318         struct tcp_md5sig_info  __rcu *md5sig_info;
 319 #endif
 320
 321         /* When the cookie options are generated and exchanged, then this
 322          * object holds a reference to them (cookie_values->kref).  Also
 323          * contains related tcp_cookie_transactions fields.
 324          */
 325         struct tcp_cookie_values  *cookie_values;
 326
 327 /* TCP fastopen related information */
 328         struct tcp_fastopen_request *fastopen_req;
 329         /* fastopen_rsk points to request_sock that resulted in this big
 330          * socket. Used to retransmit SYNACKs etc.
 331          */
 332         struct request_sock *fastopen_rsk;
 333 };
 334
 335 enum tsq_flags {
 336         TSQ_THROTTLED,
 337         TSQ_QUEUED,
 338         TCP_TSQ_DEFERRED,          /* tcp_tasklet_func() found socket was owned */
 339         TCP_WRITE_TIMER_DEFERRED,  /* tcp_write_timer() found socket was owned */
 340         TCP_DELACK_TIMER_DEFERRED, /* tcp_delack_timer() found socket was owned */
 341         TCP_MTU_REDUCED_DEFERRED,  /* tcp_v{4|6}_err() could not call
 342                                     * tcp_v{4|6}_mtu_reduced()
 343                                     */
 344 };
 345
 346 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 347 {
 348         return (struct tcp_sock *)sk;
 349 }
 350
 351 struct tcp_timewait_sock {
 352         struct inet_timewait_sock tw_sk;
 353         u32                       tw_rcv_nxt;
 354         u32                       tw_snd_nxt;
 355         u32                       tw_rcv_wnd;
 356         u32                       tw_ts_recent;
 357         long                      tw_ts_recent_stamp;
 358 #ifdef CONFIG_TCP_MD5SIG
 359         struct tcp_md5sig_key     *tw_md5_key;
 360 #endif
 361         /* Few sockets in timewait have cookies; in that case, then this
 362          * object holds a reference to them (tw_cookie_values->kref).
 363          */
 364         struct tcp_cookie_values  *tw_cookie_values;
 365 };
 366
 367 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 368 {
 369         return (struct tcp_timewait_sock *)sk;
 370 }
 371
 372 static inline bool tcp_passive_fastopen(const struct sock *sk)
 373 {
 374         return (sk->sk_state == TCP_SYN_RECV &&
 375                 tcp_sk(sk)->fastopen_rsk != NULL);
 376 }
 377
 378 static inline bool fastopen_cookie_present(struct tcp_fastopen_cookie *foc)
 379 {
 380         return foc->len != -1;
 381 }
 382
 383 extern void tcp_sock_destruct(struct sock *sk);
 384
 385 static inline int fastopen_init_queue(struct sock *sk, int backlog)
 386 {
 387         struct request_sock_queue *queue =
 388             &inet_csk(sk)->icsk_accept_queue;
 389
 390         if (queue->fastopenq == NULL) {
 391                 queue->fastopenq = kzalloc(
 392                     sizeof(struct fastopen_queue),
 393                     sk->sk_allocation);
 394                 if (queue->fastopenq == NULL)
 395                         return -ENOMEM;
 396
 397                 sk->sk_destruct = tcp_sock_destruct;
 398                 spin_lock_init(&queue->fastopenq->lock);
 399         }
 400         queue->fastopenq->max_qlen = backlog;
 401         return 0;
 402 }
 403
 404 #endif  /* _LINUX_TCP_H */