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      num_sacks;      /* Number of SACK blocks                */
  94         u16     user_mss;       /* mss requested by user in ioctl       */
  95         u16     mss_clamp;      /* Maximal mss, negotiated at connection setup */
  96 };
  97
  98 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
  99 {
 100         rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
 101         rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
 102 }
 103
 104 /* This is the max number of SACKS that we'll generate and process. It's safe
 105  * to increase this, although since:
 106  *   size = TCPOLEN_SACK_BASE_ALIGNED (4) + n * TCPOLEN_SACK_PERBLOCK (8)
 107  * only four options will fit in a standard TCP header */
 108 #define TCP_NUM_SACKS 4
 109
 110 struct tcp_cookie_values;
 111 struct tcp_request_sock_ops;
 112
 113 struct tcp_request_sock {
 114         struct inet_request_sock        req;
 115 #ifdef CONFIG_TCP_MD5SIG
 116         /* Only used by TCP MD5 Signature so far. */
 117         const struct tcp_request_sock_ops *af_specific;
 118 #endif
 119         struct sock                     *listener; /* needed for TFO */
 120         u32                             rcv_isn;
 121         u32                             snt_isn;
 122         u32                             snt_synack; /* synack sent time */
 123         u32                             rcv_nxt; /* the ack # by SYNACK. For
 124                                                   * FastOpen it's the seq#
 125                                                   * after data-in-SYN.
 126                                                   */
 127 };
 128
 129 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 130 {
 131         return (struct tcp_request_sock *)req;
 132 }
 133
 134 struct tcp_sock {
 135         /* inet_connection_sock has to be the first member of tcp_sock */
 136         struct inet_connection_sock     inet_conn;
 137         u16     tcp_header_len; /* Bytes of tcp header to send          */
 138         u16     xmit_size_goal_segs; /* Goal for segmenting output packets */
 139
 140 /*
 141  *      Header prediction flags
 142  *      0x5?10 << 16 + snd_wnd in net byte order
 143  */
 144         __be32  pred_flags;
 145
 146 /*
 147  *      RFC793 variables by their proper names. This means you can
 148  *      read the code and the spec side by side (and laugh ...)
 149  *      See RFC793 and RFC1122. The RFC writes these in capitals.
 150  */
 151         u32     rcv_nxt;        /* What we want to receive next         */
 152         u32     copied_seq;     /* Head of yet unread data              */
 153         u32     rcv_wup;        /* rcv_nxt on last window update sent   */
 154         u32     snd_nxt;        /* Next sequence we send                */
 155
 156         u32     snd_una;        /* First byte we want an ack for        */
 157         u32     snd_sml;        /* Last byte of the most recently transmitted small packet */
 158         u32     rcv_tstamp;     /* timestamp of last received ACK (for keepalives) */
 159         u32     lsndtime;       /* timestamp of last sent data packet (for restart window) */
 160
 161         u32     tsoffset;       /* timestamp offset */
 162
 163         struct list_head tsq_node; /* anchor in tsq_tasklet.head list */
 164         unsigned long   tsq_flags;
 165
 166         /* Data for direct copy to user */
 167         struct {
 168                 struct sk_buff_head     prequeue;
 169                 struct task_struct      *task;
 170                 struct iovec            *iov;
 171                 int                     memory;
 172                 int                     len;
 173 #ifdef CONFIG_NET_DMA
 174                 /* members for async copy */
 175                 struct dma_chan         *dma_chan;
 176                 int                     wakeup;
 177                 struct dma_pinned_list  *pinned_list;
 178                 dma_cookie_t            dma_cookie;
 179 #endif
 180         } ucopy;
 181
 182         u32     snd_wl1;        /* Sequence for window update           */
 183         u32     snd_wnd;        /* The window we expect to receive      */
 184         u32     max_window;     /* Maximal window ever seen from peer   */
 185         u32     mss_cache;      /* Cached effective mss, not including SACKS */
 186
 187         u32     window_clamp;   /* Maximal window to advertise          */
 188         u32     rcv_ssthresh;   /* Current window clamp                 */
 189
 190         u16     advmss;         /* Advertised MSS                       */
 191         u8      unused;
 192         u8      nonagle     : 4,/* Disable Nagle algorithm?             */
 193                 thin_lto    : 1,/* Use linear timeouts for thin streams */
 194                 thin_dupack : 1,/* Fast retransmit on first dupack      */
 195                 repair      : 1,
 196                 frto        : 1;/* F-RTO (RFC5682) activated in CA_Loss */
 197         u8      repair_queue;
 198         u8      do_early_retrans:1,/* Enable RFC5827 early-retransmit  */
 199                 syn_data:1,     /* SYN includes data */
 200                 syn_fastopen:1, /* SYN includes Fast Open option */
 201                 syn_data_acked:1;/* data in SYN is acked by SYN-ACK */
 202         u32     tlp_high_seq;   /* snd_nxt at the time of TLP retransmit. */
 203
 204 /* RTT measurement */
 205         u32     srtt;           /* smoothed round trip time << 3        */
 206         u32     mdev;           /* medium deviation                     */
 207         u32     mdev_max;       /* maximal mdev for the last rtt period */
 208         u32     rttvar;         /* smoothed mdev_max                    */
 209         u32     rtt_seq;        /* sequence number to update rttvar     */
 210
 211         u32     packets_out;    /* Packets which are "in flight"        */
 212         u32     retrans_out;    /* Retransmitted packets out            */
 213
 214         u16     urg_data;       /* Saved octet of OOB data and control flags */
 215         u8      ecn_flags;      /* ECN status bits.                     */
 216         u8      reordering;     /* Packet reordering metric.            */
 217         u32     snd_up;         /* Urgent pointer               */
 218
 219         u8      keepalive_probes; /* num of allowed keep alive probes   */
 220 /*
 221  *      Options received (usually on last packet, some only on SYN packets).
 222  */
 223         struct tcp_options_received rx_opt;
 224
 225 /*
 226  *      Slow start and congestion control (see also Nagle, and Karn & Partridge)
 227  */
 228         u32     snd_ssthresh;   /* Slow start size threshold            */
 229         u32     snd_cwnd;       /* Sending congestion window            */
 230         u32     snd_cwnd_cnt;   /* Linear increase counter              */
 231         u32     snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 232         u32     snd_cwnd_used;
 233         u32     snd_cwnd_stamp;
 234         u32     prior_cwnd;     /* Congestion window at start of Recovery. */
 235         u32     prr_delivered;  /* Number of newly delivered packets to
 236                                  * receiver in Recovery. */
 237         u32     prr_out;        /* Total number of pkts sent during Recovery. */
 238
 239         u32     rcv_wnd;        /* Current receiver window              */
 240         u32     write_seq;      /* Tail(+1) of data held in tcp send buffer */
 241         u32     pushed_seq;     /* Last pushed seq, required to talk to windows */
 242         u32     lost_out;       /* Lost packets                 */
 243         u32     sacked_out;     /* SACK'd packets                       */
 244         u32     fackets_out;    /* FACK'd packets                       */
 245         u32     tso_deferred;
 246
 247         /* from STCP, retrans queue hinting */
 248         struct sk_buff* lost_skb_hint;
 249         struct sk_buff *scoreboard_skb_hint;
 250         struct sk_buff *retransmit_skb_hint;
 251
 252         struct sk_buff_head     out_of_order_queue; /* Out of order segments go here */
 253
 254         /* SACKs data, these 2 need to be together (see tcp_options_write) */
 255         struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 256         struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 257
 258         struct tcp_sack_block recv_sack_cache[4];
 259
 260         struct sk_buff *highest_sack;   /* skb just after the highest
 261                                          * skb with SACKed bit set
 262                                          * (validity guaranteed only if
 263                                          * sacked_out > 0)
 264                                          */
 265
 266         int     lost_cnt_hint;
 267         u32     retransmit_high;        /* L-bits may be on up to this seqno */
 268
 269         u32     lost_retrans_low;       /* Sent seq after any rxmit (lowest) */
 270
 271         u32     prior_ssthresh; /* ssthresh saved at recovery start     */
 272         u32     high_seq;       /* snd_nxt at onset of congestion       */
 273
 274         u32     retrans_stamp;  /* Timestamp of the last retransmit,
 275                                  * also used in SYN-SENT to remember stamp of
 276                                  * the first SYN. */
 277         u32     undo_marker;    /* tracking retrans started here. */
 278         int     undo_retrans;   /* number of undoable retransmissions. */
 279         u32     total_retrans;  /* Total retransmits for entire connection */
 280
 281         u32     urg_seq;        /* Seq of received urgent pointer */
 282         unsigned int            keepalive_time;   /* time before keep alive takes place */
 283         unsigned int            keepalive_intvl;  /* time interval between keep alive probes */
 284
 285         int                     linger2;
 286
 287 /* Receiver side RTT estimation */
 288         struct {
 289                 u32     rtt;
 290                 u32     seq;
 291                 u32     time;
 292         } rcv_rtt_est;
 293
 294 /* Receiver queue space */
 295         struct {
 296                 int     space;
 297                 u32     seq;
 298                 u32     time;
 299         } rcvq_space;
 300
 301 /* TCP-specific MTU probe information. */
 302         struct {
 303                 u32               probe_seq_start;
 304                 u32               probe_seq_end;
 305         } mtu_probe;
 306         u32     mtu_info; /* We received an ICMP_FRAG_NEEDED / ICMPV6_PKT_TOOBIG
 307                            * while socket was owned by user.
 308                            */
 309
 310 #ifdef CONFIG_TCP_MD5SIG
 311 /* TCP AF-Specific parts; only used by MD5 Signature support so far */
 312         const struct tcp_sock_af_ops    *af_specific;
 313
 314 /* TCP MD5 Signature Option information */
 315         struct tcp_md5sig_info  __rcu *md5sig_info;
 316 #endif
 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
 326 enum 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
 337 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 338 {
 339         return (struct tcp_sock *)sk;
 340 }
 341
 342 struct 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_offset;
 348         u32                       tw_ts_recent;
 349         long                      tw_ts_recent_stamp;
 350 #ifdef CONFIG_TCP_MD5SIG
 351         struct tcp_md5sig_key     *tw_md5_key;
 352 #endif
 353 };
 354
 355 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
 356 {
 357         return (struct tcp_timewait_sock *)sk;
 358 }
 359
 360 static inline bool tcp_passive_fastopen(const struct sock *sk)
 361 {
 362         return (sk->sk_state == TCP_SYN_RECV &&
 363                 tcp_sk(sk)->fastopen_rsk != NULL);
 364 }
 365
 366 static inline bool fastopen_cookie_present(struct tcp_fastopen_cookie *foc)
 367 {
 368         return foc->len != -1;
 369 }
 370
 371 extern void tcp_sock_destruct(struct sock *sk);
 372
 373 static inline int fastopen_init_queue(struct sock *sk, int backlog)
 374 {
 375         struct request_sock_queue *queue =
 376             &inet_csk(sk)->icsk_accept_queue;
 377
 378         if (queue->fastopenq == NULL) {
 379                 queue->fastopenq = kzalloc(
 380                     sizeof(struct fastopen_queue),
 381                     sk->sk_allocation);
 382                 if (queue->fastopenq == NULL)
 383                         return -ENOMEM;
 384
 385                 sk->sk_destruct = tcp_sock_destruct;
 386                 spin_lock_init(&queue->fastopenq->lock);
 387         }
 388         queue->fastopenq->max_qlen = backlog;
 389         return 0;
 390 }
 391
 392 #endif  /* _LINUX_TCP_H */