bsd/netinet/tcp_var.h

   1 /*
   2  * Copyright (c) 2000-2016 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /*
  29  * Copyright (c) 1982, 1986, 1993, 1994, 1995
  30  *      The Regents of the University of California.  All rights reserved.
  31  *
  32  * Redistribution and use in source and binary forms, with or without
  33  * modification, are permitted provided that the following conditions
  34  * are met:
  35  * 1. Redistributions of source code must retain the above copyright
  36  *    notice, this list of conditions and the following disclaimer.
  37  * 2. Redistributions in binary form must reproduce the above copyright
  38  *    notice, this list of conditions and the following disclaimer in the
  39  *    documentation and/or other materials provided with the distribution.
  40  * 3. All advertising materials mentioning features or use of this software
  41  *    must display the following acknowledgement:
  42  *      This product includes software developed by the University of
  43  *      California, Berkeley and its contributors.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)tcp_var.h   8.4 (Berkeley) 5/24/95
  61  * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.8 2001/08/22 00:59:13 silby Exp $
  62  */
  63
  64 #ifndef _NETINET_TCP_VAR_H_
  65 #define _NETINET_TCP_VAR_H_
  66 #include <sys/types.h>
  67 #include <sys/appleapiopts.h>
  68 #include <sys/queue.h>
  69 #include <netinet/in_pcb.h>
  70 #include <netinet/tcp.h>
  71 #include <netinet/tcp_timer.h>
  72
  73 #if defined(__LP64__)
  74 #define _TCPCB_PTR(x)                   u_int32_t
  75 #define _TCPCB_LIST_HEAD(name, type)    \
  76 struct name {                           \
  77         u_int32_t       lh_first;       \
  78 }
  79 #else
  80 #define _TCPCB_PTR(x)                   x
  81 #define _TCPCB_LIST_HEAD(name, type)    LIST_HEAD(name, type)
  82 #endif
  83
  84 #ifdef KERNEL_PRIVATE
  85
  86 #define TCP_RETRANSHZ   1000    /* granularity of TCP timestamps, 1ms */
  87 /* Minimum time quantum within which the timers are coalesced */
  88 #define TCP_TIMER_10MS_QUANTUM  (TCP_RETRANSHZ/100) /* every 10ms */
  89 #define TCP_TIMER_100MS_QUANTUM   (TCP_RETRANSHZ/10) /* every 100ms */
  90 #define TCP_TIMER_500MS_QUANTUM   (TCP_RETRANSHZ/2) /* every 500ms */
  91
  92 #define TCP_RETRANSHZ_TO_USEC 1000
  93
  94 #define N_TIME_WAIT_SLOTS   128         /* must be power of 2 */
  95
  96 /* Always allow at least 16 packets worth of recv window when adjusting
  97  * recv window using inter-packet arrival jitter.
  98  */
  99 #define MIN_IAJ_WIN 16
 100
 101 /* A variation in delay of this many milliseconds is tolerable. This limit has to
 102  * be low but greater than zero. We also use standard deviation on jitter to adjust
 103  * this limit for different link and connection types.
 104  */
 105 #define ALLOWED_IAJ 5
 106
 107 /* Ignore the first few packets on a connection until the ACK clock gets going
 108  */
 109 #define IAJ_IGNORE_PKTCNT 40
 110
 111 /* Let the accumulated IAJ value increase by this threshold at most. This limit
 112  * will control how many ALLOWED_IAJ measurements a receiver will have to see
 113  * before opening the receive window
 114  */
 115 #define ACC_IAJ_HIGH_THRESH 100
 116
 117 /* When accumulated IAJ reaches this value, the receiver starts to react by
 118  * closing the window
 119  */
 120 #define ACC_IAJ_REACT_LIMIT 200
 121
 122 /* If the number of small packets (smaller than IAJ packet size) seen on a
 123  * connection is more than this threshold, reset the size and learn it again.
 124  * This is needed because the sender might send smaller segments after PMTU
 125  * discovery and the receiver has to learn the new size.
 126  */
 127 #define RESET_IAJ_SIZE_THRESH 20
 128
 129 /*
 130  * Adaptive timeout is a read/write timeout specified by the application to
 131  * get a socket event when the transport layer detects a stall in data
 132  * transfer. The value specified is the number of probes that can be sent
 133  * to the peer before generating an event. Since it is not specified as
 134  * a time value, the timeout will adjust based on the RTT seen on the link.
 135  * The timeout will start only when there is an indication that the read/write
 136  * operation is not making progress.
 137  *
 138  * If a write operation stalls, the probe will be retransmission of data.
 139  * If a read operation stalls, the probe will be a keep-alive packet.
 140  *
 141  * The maximum value of adaptive timeout is set to 10 which will allow
 142  * transmission of enough number of probes to the peer.
 143  */
 144 #define TCP_ADAPTIVE_TIMEOUT_MAX        10
 145
 146 #define TCP_CONNECTIVITY_PROBES_MAX     5
 147
 148 /*
 149  * Kernel variables for tcp.
 150  */
 151
 152 /* TCP segment queue entry */
 153 struct tseg_qent {
 154         LIST_ENTRY(tseg_qent) tqe_q;
 155         int     tqe_len;                /* TCP segment data length */
 156         struct  tcphdr *tqe_th;         /* a pointer to tcp header */
 157         struct  mbuf    *tqe_m;         /* mbuf contains packet */
 158 };
 159 LIST_HEAD(tsegqe_head, tseg_qent);
 160
 161 struct sackblk {
 162         tcp_seq start;          /* start seq no. of sack block */
 163         tcp_seq end;            /* end seq no. */
 164 };
 165
 166 struct sackhole {
 167         tcp_seq start;          /* start seq no. of hole */
 168         tcp_seq end;            /* end seq no. */
 169         tcp_seq rxmit;          /* next seq. no in hole to be retransmitted */
 170         u_int32_t rxmit_start;  /* timestamp of first retransmission */
 171         TAILQ_ENTRY(sackhole) scblink;  /* scoreboard linkage */
 172 };
 173
 174 struct sackhint {
 175         struct sackhole *nexthole;
 176         int     sack_bytes_rexmit;
 177 };
 178
 179 struct tcp_rxt_seg {
 180         tcp_seq rx_start;
 181         tcp_seq rx_end;
 182         u_int16_t rx_count;
 183         u_int16_t rx_flags;
 184 #define TCP_RXT_SPURIOUS        0x1     /* received DSACK notification */
 185 #define TCP_RXT_DSACK_FOR_TLP   0x2
 186         SLIST_ENTRY(tcp_rxt_seg) rx_link;
 187 };
 188
 189 struct tcp_notify_ack_marker {
 190         tcp_seq notify_snd_una; /* Notify when snd_una crosses this seq */
 191         tcp_notify_ack_id_t notify_id;
 192         SLIST_ENTRY(tcp_notify_ack_marker) notify_next;
 193 };
 194
 195 struct tcptemp {
 196         u_char  tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
 197         struct  tcphdr tt_t;
 198 };
 199
 200 struct bwmeas {
 201         tcp_seq bw_start;               /* start of bw measurement */
 202         uint32_t bw_ts;         /* timestamp when bw measurement started */
 203         uint32_t bw_size;               /* burst size in bytes for this bw measurement */
 204         uint32_t bw_minsizepkts;        /* Min burst size as segments */
 205         uint32_t bw_maxsizepkts;        /* Max burst size as segments */
 206         uint32_t bw_minsize;    /* Min size in bytes */
 207         uint32_t bw_maxsize;    /* Max size in bytes */
 208         uint32_t bw_sndbw;              /* Measured send bw */
 209 };
 210
 211 /* MPTCP Data sequence map entry */
 212 struct mpt_dsn_map {
 213         uint64_t                mpt_dsn;        /* data seq num recvd */
 214         uint32_t                mpt_sseq;       /* relative subflow # */
 215         uint16_t                mpt_len;        /* length of mapping */
 216         uint16_t                mpt_csum;       /* checksum value if on */
 217 };
 218 #define tcp6cb          tcpcb  /* for KAME src sync over BSD*'s */
 219
 220 struct tcp_ccstate {
 221         union {
 222                 struct tcp_cubic_state {
 223                         u_int32_t tc_last_max;  /* cwnd at last loss */
 224                         u_int32_t tc_epoch_start; /* TS of last loss */
 225                         u_int32_t tc_origin_point; /* window at the start of an epoch */
 226                         u_int32_t tc_tcp_win; /* computed tcp win */
 227                         u_int32_t tc_tcp_bytes_acked; /* bytes acked */
 228                         u_int32_t tc_target_win; /* cubic target win */
 229                         u_int32_t tc_avg_lastmax; /* Average of last max */
 230                         u_int32_t tc_mean_deviation; /* Mean absolute deviation */
 231                         float     tc_epoch_period; /* K parameter */
 232                 } _cubic_state_;
 233 #define cub_last_max __u__._cubic_state_.tc_last_max
 234 #define cub_epoch_start __u__._cubic_state_.tc_epoch_start
 235 #define cub_origin_point __u__._cubic_state_.tc_origin_point
 236 #define cub_tcp_win __u__._cubic_state_.tc_tcp_win
 237 #define cub_tcp_bytes_acked __u__._cubic_state_.tc_tcp_bytes_acked
 238 #define cub_epoch_period __u__._cubic_state_.tc_epoch_period
 239 #define cub_target_win __u__._cubic_state_.tc_target_win
 240 #define cub_avg_lastmax __u__._cubic_state_.tc_avg_lastmax
 241 #define cub_mean_dev __u__._cubic_state_.tc_mean_deviation
 242         } __u__;
 243 };
 244
 245 /*
 246  * Tcp control block, one per tcp; fields:
 247  * Organized for 16 byte cacheline efficiency.
 248  */
 249 struct tcpcb {
 250         struct  tsegqe_head t_segq;
 251         int     t_dupacks;              /* consecutive dup acks recd */
 252         int     t_state;                /* state of this connection */
 253         uint32_t t_timer[TCPT_NTIMERS]; /* tcp timers */
 254         struct tcptimerentry tentry;    /* entry in timer list */
 255
 256         struct  inpcb *t_inpcb;         /* back pointer to internet pcb */
 257         uint32_t        t_flags;
 258 #define TF_ACKNOW       0x00001         /* ack peer immediately */
 259 #define TF_DELACK       0x00002         /* ack, but try to delay it */
 260 #define TF_NODELAY      0x00004         /* don't delay packets to coalesce */
 261 #define TF_NOOPT        0x00008         /* don't use tcp options */
 262 #define TF_SENTFIN      0x00010         /* have sent FIN */
 263 #define TF_REQ_SCALE    0x00020         /* have/will request window scaling */
 264 #define TF_RCVD_SCALE   0x00040         /* other side has requested scaling */
 265 #define TF_REQ_TSTMP    0x00080         /* have/will request timestamps */
 266 #define TF_RCVD_TSTMP   0x00100         /* a timestamp was received in SYN */
 267 #define TF_SACK_PERMIT  0x00200         /* other side said I could SACK */
 268 #define TF_NEEDSYN      0x00400         /* send SYN (implicit state) */
 269 #define TF_NEEDFIN      0x00800         /* send FIN (implicit state) */
 270 #define TF_NOPUSH       0x01000         /* don't push */
 271 #define TF_REQ_CC       0x02000         /* have/will request CC */
 272 #define TF_RCVD_CC      0x04000         /* a CC was received in SYN */
 273 #define TF_SENDCCNEW    0x08000         /* Unused */
 274 #define TF_MORETOCOME   0x10000         /* More data to be appended to sock */
 275 #define TF_LOCAL        0x20000         /* connection to a host on local link */
 276 #define TF_RXWIN0SENT   0x40000         /* sent a receiver win 0 in response */
 277 #define TF_SLOWLINK     0x80000         /* route is a on a modem speed link */
 278 #define TF_LASTIDLE     0x100000        /* connection was previously idle */
 279 #define TF_FASTRECOVERY 0x200000        /* in NewReno Fast Recovery */
 280 #define TF_WASFRECOVERY 0x400000        /* was in NewReno Fast Recovery */
 281 #define TF_SIGNATURE    0x800000        /* require MD5 digests (RFC2385) */
 282 #define TF_MAXSEGSNT    0x1000000       /* last segment sent was a full segment */
 283 #define TF_STREAMING_ON 0x2000000       /* Receiver detected streaming */
 284 #define TF_PMTUD        0x4000000       /* Perform Path MTU Discovery for this connection */
 285 #define TF_CLOSING      0x8000000       /* pending tcp close */
 286 #define TF_TSO          0x10000000      /* TCP Segment Offloading is enable on this connection */
 287 #define TF_BLACKHOLE    0x20000000      /* Path MTU Discovery Black Hole detection */
 288 #define TF_TIMER_ONLIST 0x40000000      /* pcb is on tcp_timer_list */
 289 #define TF_STRETCHACK   0x80000000      /* receiver is going to delay acks */
 290
 291         tcp_seq snd_una;                /* send unacknowledged */
 292         tcp_seq snd_max;                /* highest sequence number sent;
 293                                          * used to recognize retransmits
 294                                          */
 295         tcp_seq snd_nxt;                /* send next */
 296         tcp_seq snd_up;                 /* send urgent pointer */
 297
 298         tcp_seq snd_wl1;                /* window update seg seq number */
 299         tcp_seq snd_wl2;                /* window update seg ack number */
 300         tcp_seq iss;                    /* initial send sequence number */
 301         tcp_seq irs;                    /* initial receive sequence number */
 302
 303         tcp_seq rcv_nxt;                /* receive next */
 304         tcp_seq rcv_adv;                /* advertised window */
 305         u_int32_t       rcv_wnd;                /* receive window */
 306         tcp_seq rcv_up;                 /* receive urgent pointer */
 307
 308         u_int32_t       snd_wnd;        /* send window */
 309         u_int32_t       snd_cwnd;       /* congestion-controlled window */
 310         u_int32_t       snd_ssthresh;   /* snd_cwnd size threshold for
 311                                          * for slow start exponential to
 312                                          * linear switch
 313                                          */
 314         tcp_seq snd_recover;            /* for use in NewReno Fast Recovery */
 315
 316         u_int32_t       t_maxopd;       /* mss plus options */
 317         u_int32_t       t_rcvtime;      /* time at which a packet was received */
 318         u_int32_t       t_starttime;    /* time connection was established */
 319         int     t_rtttime;              /* tcp clock when rtt calculation was started */
 320         tcp_seq t_rtseq;                /* sequence number being timed */
 321
 322         u_int32_t rfbuf_ts;             /* recv buffer autoscaling timestamp */
 323         u_int32_t rfbuf_cnt;            /* recv buffer autoscaling byte count */
 324
 325         int     t_rxtcur;               /* current retransmit value (ticks) */
 326         u_int   t_maxseg;               /* maximum segment size */
 327         int     t_srtt;                 /* smoothed round-trip time */
 328         int     t_rttvar;               /* variance in round-trip time */
 329
 330         u_int64_t       t_accsleep_ms;  /* accumulated sleep time since last boot */
 331         u_int16_t       t_reassqlen;    /* length of reassembly queue */
 332         u_int16_t       t_rxtshift;     /* log(2) of rexmt exp. backoff */
 333         u_int32_t       t_rttmin;       /* minimum rtt allowed */
 334         u_int32_t       t_rttbest;      /* best rtt we've seen */
 335         u_int32_t       t_rttcur;       /* most recent value of rtt */
 336         u_int32_t       t_rttupdated;   /* number of times rtt sampled */
 337         u_int32_t       t_rxt_conndroptime;     /* retxmt conn gets dropped after this time, when set */
 338         u_int32_t       t_rxtstart;     /* time at which retransmission started */
 339         u_int32_t       max_sndwnd;     /* largest window peer has offered */
 340
 341         int     t_softerror;            /* possible error not yet reported */
 342 /* out-of-band data */
 343         char    t_oobflags;             /* have some */
 344         char    t_iobc;                 /* input character */
 345 #define TCPOOB_HAVEDATA 0x01
 346 #define TCPOOB_HADDATA  0x02
 347 /* RFC 1323 variables */
 348         u_int8_t        snd_scale;      /* window scaling for send window */
 349         u_int8_t        rcv_scale;      /* window scaling for recv window */
 350         u_int8_t        request_r_scale; /* pending window scaling */
 351         u_int8_t        requested_s_scale;
 352         u_int8_t        tcp_cc_index;   /* index of congestion control algorithm */
 353         u_int8_t        t_adaptive_rtimo;       /* Read timeout used as a multiple of RTT */
 354         u_int8_t        t_adaptive_wtimo;       /* Write timeout used as a multiple of RTT */
 355         u_int8_t        t_stretchack_delayed;   /* stretch ack delayed */
 356
 357 /* State for limiting early retransmits when SACK is not enabled */
 358         u_int16_t       t_early_rexmt_count; /* count of early rexmts */
 359         u_int32_t       t_early_rexmt_win; /* window for limiting early rexmts */
 360
 361         u_int32_t       ts_recent;      /* timestamp echo data */
 362
 363         u_int32_t       ts_recent_age;  /* when last updated */
 364         tcp_seq last_ack_sent;
 365 /* RFC 3465 variables */
 366         u_int32_t       t_bytes_acked;  /* ABC "bytes_acked" parameter */
 367
 368         int     t_lastchain;            /* amount of packets chained last time around */
 369         u_int16_t       t_unacksegs;    /* received but unacked segments for delaying acks */
 370         u_int8_t        t_rexmtthresh;  /* duplicate ack threshold for entering fast recovery */
 371         u_int8_t        t_rtimo_probes; /* number of adaptive rtimo probes sent */
 372         u_int32_t       t_persist_timeout; /* ZWP persistence limit as set by PERSIST_TIMEOUT */
 373         u_int32_t       t_persist_stop;         /* persistence limit deadline if triggered by ZWP */
 374         u_int32_t       t_notsent_lowat;        /* Low water for not sent data */
 375
 376 /* Receiver state for stretch-ack algorithm */
 377         u_int32_t       rcv_unackwin;   /* to measure win for stretching acks */
 378         u_int32_t       rcv_by_unackwin; /* bytes seen during the last ack-stretching win */
 379         u_int32_t       rcv_nostrack_ts; /* timestamp when stretch ack was disabled automatically */
 380         u_int32_t       rcv_nostrack_pkts; /* pkts received since strech ack was disabled */
 381         u_int16_t       rcv_waitforss;  /* wait for packets during slow-start */
 382
 383 /* ECN stats */
 384         u_int16_t       ecn_flags;
 385 #define TE_SETUPSENT            0x0001  /* Indicate we have sent ECN-SETUP SYN or SYN-ACK */
 386 #define TE_SETUPRECEIVED        0x0002  /* Indicate we have received ECN-SETUP SYN or SYN-ACK */
 387 #define TE_SENDIPECT            0x0004  /* Indicate we haven't sent or received non-ECN-setup SYN or SYN-ACK */
 388 #define TE_SENDCWR              0x0008  /* Indicate that the next non-retransmit should have the TCP CWR flag set */
 389 #define TE_SENDECE              0x0010  /* Indicate that the next packet should have the TCP ECE flag set */
 390 #define TE_INRECOVERY           0x0020  /* connection entered recovery after receiving ECE */
 391 #define TE_RECV_ECN_CE          0x0040  /* Received IPTOS_ECN_CE marking atleast once */
 392 #define TE_RECV_ECN_ECE         0x0080  /* Received ECE marking atleast once */
 393 #define TE_LOST_SYN             0x0100  /* Lost SYN with ECN setup */
 394 #define TE_LOST_SYNACK          0x0200  /* Lost SYN-ACK with ECN setup */
 395 #define TE_ECN_MODE_ENABLE      0x0400  /* Option ECN mode set to enable */
 396 #define TE_ECN_MODE_DISABLE     0x0800  /* Option ECN mode set to disable */
 397 #define TE_ENABLE_ECN           0x1000  /* Enable negotiation of ECN */
 398 #define TE_ECN_ON               (TE_SETUPSENT | TE_SETUPRECEIVED) /* Indicate ECN was successfully negotiated on a connection) */
 399 #define TE_CEHEURI_SET          0x2000 /* We did our CE-probing at the beginning */
 400 #define TE_CLIENT_SETUP         0x4000  /* setup from client side */
 401
 402         u_int32_t       t_ecn_recv_ce;  /* Received CE from the network */
 403         u_int32_t       t_ecn_recv_cwr; /* Packets received with CWR */
 404
 405 /* state for bad retransmit recovery */
 406         u_int32_t       snd_cwnd_prev;  /* cwnd prior to retransmit */
 407         u_int32_t       snd_ssthresh_prev; /* ssthresh prior to retransmit */
 408         tcp_seq snd_recover_prev;       /* snd_recover prior to retransmit */
 409         int     t_srtt_prev;            /* srtt prior to retransmit */
 410         int     t_rttvar_prev;          /* rttvar prior to retransmit */
 411         u_int32_t       t_badrexmt_time; /* bad rexmt detection time */
 412
 413 /* Packet reordering metric */
 414         u_int16_t       t_reorderwin; /* Reordering late time offset */
 415
 416 /* SACK related state */
 417         int16_t snd_numholes;           /* number of holes seen by sender */
 418         tcp_seq sack_newdata;           /* New data xmitted in this recovery
 419                                            episode starts at this seq number */
 420         TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
 421                                                 /* SACK scoreboard (sorted) */
 422         tcp_seq snd_fack;               /* last seq number(+1) sack'd by rcv'r*/
 423         int     rcv_numsacks;           /* # distinct sack blks present */
 424         struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
 425         struct sackhint sackhint;       /* SACK scoreboard hint */
 426
 427         struct mbuf     *t_pktlist_head; /* First packet in transmit chain */
 428         struct mbuf     *t_pktlist_tail; /* Last packet in transmit chain */
 429         u_int32_t       t_pktlist_sentlen; /* total bytes in transmit chain */
 430
 431         u_int32_t       t_keepidle;     /* keepalive idle timer (override global if > 0) */
 432         u_int32_t       t_keepinit;     /* connection timeout, i.e. idle time
 433                                         in SYN_SENT or SYN_RECV state */
 434         u_int32_t       t_keepintvl;    /* interval between keepalives */
 435         u_int32_t       t_keepcnt;      /* number of keepalives before close */
 436
 437         u_int32_t       tso_max_segment_size;   /* TSO maximum segment unit for NIC */
 438         u_int16_t       t_pmtud_lastseg_size;   /* size of the last sent segment */
 439         u_int16_t       t_pmtud_saved_maxopd;   /* MSS saved before performing PMTU-D BlackHole detection */
 440         u_int32_t       t_pmtud_start_ts;       /* Time of PMTUD blackhole detection */
 441
 442         struct
 443         {
 444                 u_int32_t       rxduplicatebytes;
 445                 u_int32_t       rxoutoforderbytes;
 446                 u_int32_t       txretransmitbytes;
 447                 u_int8_t        synrxtshift;
 448                 u_int8_t        unused;
 449                 u_int16_t       unused_pad_to_8;
 450                 u_int32_t       rxmitpkts;
 451         } t_stat;
 452         u_int8_t        t_notify_ack_count;
 453         u_int8_t        t_ecn_recv_ce_pkt; /* Received packet with CE-bit set (independent from last_ack_sent) */
 454         u_int16_t       t_cached_maxopd; /* default for MSS adjustment using link status report */
 455
 456         uint32_t        bg_ssthresh;            /* Slow start threshold until delay increases */
 457         uint32_t        t_flagsext;             /* Another field to accommodate more flags */
 458 #define TF_RXTFINDROP           0x1                     /* Drop conn after retransmitting FIN 3 times */
 459 #define TF_RCVUNACK_WAITSS      0x2             /* set when the receiver should not stretch acks */
 460 #define TF_BWMEAS_INPROGRESS    0x4             /* Indicate BW meas is happening */
 461 #define TF_MEASURESNDBW         0x8             /* Measure send bw on this connection */
 462 #define TF_LRO_OFFLOADED        0x10            /* Connection LRO offloaded */
 463 #define TF_SACK_ENABLE          0x20            /* SACK is enabled */
 464 #define TF_RECOMPUTE_RTT        0x40            /* recompute RTT after spurious retransmit */
 465 #define TF_DETECT_READSTALL     0x80            /* Used to detect a stall during read operation */
 466 #define TF_RECV_THROTTLE        0x100           /* Input throttling active */
 467 #define TF_NOSTRETCHACK         0x200           /* ack every other packet */
 468 #define TF_NOTIMEWAIT           0x800           /* Avoid going into time-wait */
 469 #define TF_SENT_TLPROBE         0x1000          /* Sent data in PTO */
 470 #define TF_PKTS_REORDERED       0x2000          /* Detected reordering */
 471 #define TF_DELAY_RECOVERY       0x4000          /* delay fast recovery */
 472 #define TF_FORCE                0x8000          /* force 1 byte out */
 473 #define TF_DISABLE_STRETCHACK   0x10000         /* auto-disable stretch ack */
 474 #define TF_NOBLACKHOLE_DETECTION 0x20000        /* Disable PMTU blackhole detection */
 475 #define TF_DISABLE_DSACK        0x40000         /* Ignore DSACK due to n/w duplication */
 476 #define TF_RESCUE_RXT           0x80000         /* SACK rescue retransmit */
 477 #define TF_CWND_NONVALIDATED    0x100000        /* cwnd non validated */
 478 #define TF_PROBING              0x200000        /* Trigger probe timeout */
 479 #define TF_FASTOPEN             0x400000        /* TCP Fastopen is enabled */
 480 #define TF_REASS_INPROG         0x800000        /* Reassembly is in progress */
 481
 482 #if TRAFFIC_MGT
 483         /* Inter-arrival jitter related state */
 484         uint32_t        iaj_rcv_ts;             /* tcp clock when the first packet was received */
 485         uint16_t        iaj_size;               /* Size of packet for iaj measurement */
 486         uint8_t         iaj_small_pkt;          /* Count of packets smaller than iaj_size */
 487         uint8_t         t_pipeack_ind;          /* index for next pipeack sample */
 488         uint16_t        iaj_pktcnt;             /* packet count, to avoid throttling initially */
 489         uint16_t        acc_iaj;                /* Accumulated iaj */
 490         tcp_seq         iaj_rwintop;            /* recent max advertised window */
 491         uint32_t        avg_iaj;                /* Mean */
 492         uint32_t        std_dev_iaj;            /* Standard deviation */
 493 #endif /* TRAFFIC_MGT */
 494         struct bwmeas   *t_bwmeas;              /* State for bandwidth measurement */
 495         uint32_t        t_lropktlen;            /* Bytes in a LRO frame */
 496         tcp_seq         t_idleat;               /* rcv_nxt at idle time */
 497         TAILQ_ENTRY(tcpcb) t_twentry;           /* link for time wait queue */
 498         struct tcp_ccstate      *t_ccstate;     /* congestion control related state */
 499 /* Tail loss probe related state */
 500         tcp_seq         t_tlphighrxt;           /* snd_nxt after PTO */
 501         u_int32_t       t_tlpstart;             /* timestamp at PTO */
 502 /* DSACK data receiver state */
 503         tcp_seq         t_dsack_lseq;           /* DSACK left sequence */
 504         tcp_seq         t_dsack_rseq;           /* DSACK right sequence */
 505 /* DSACK data sender state */
 506         SLIST_HEAD(tcp_rxt_seghead, tcp_rxt_seg) t_rxt_segments;
 507         tcp_seq         t_dsack_lastuna;        /* snd_una when last recovery episode started */
 508 /* state for congestion window validation (draft-ietf-tcpm-newcwv-07) */
 509 #define TCP_PIPEACK_SAMPLE_COUNT        3
 510         u_int32_t       t_pipeack_sample[TCP_PIPEACK_SAMPLE_COUNT];     /* pipeack, bytes acked within RTT */
 511         tcp_seq         t_pipeack_lastuna; /* una when pipeack measurement started */
 512         u_int32_t       t_pipeack;
 513         u_int32_t       t_lossflightsize;
 514
 515 #if MPTCP
 516         u_int32_t       t_mpflags;              /* flags for multipath TCP */
 517
 518 #define TMPF_PREESTABLISHED     0x00000001 /* conn in pre-established state */
 519 #define TMPF_SENT_KEYS          0x00000002 /* indicates that keys were sent */
 520 #define TMPF_MPTCP_TRUE         0x00000004 /* negotiated MPTCP successfully */
 521 #define TMPF_MPTCP_RCVD_KEY     0x00000008 /* state for 3-way handshake */
 522 #define TMPF_SND_MPPRIO         0x00000010 /* send priority of subflow */
 523 #define TMPF_SND_REM_ADDR       0x00000020 /* initiate address removal */
 524 #define TMPF_RCVD_DACK          0x00000040 /* received a data-ack */
 525 #define TMPF_JOINED_FLOW        0x00000080 /* Indicates additional flow */
 526 #define TMPF_BACKUP_PATH        0x00000100 /* Indicates backup path */
 527 #define TMPF_MPTCP_ACKNOW       0x00000200 /* Send Data ACK */
 528 #define TMPF_SEND_DSN           0x00000400 /* Send DSN mapping */
 529 #define TMPF_SEND_DFIN          0x00000800 /* Send Data FIN */
 530 #define TMPF_RECV_DFIN          0x00001000 /* Recv Data FIN */
 531 #define TMPF_SENT_JOIN          0x00002000 /* Sent Join */
 532 #define TMPF_RECVD_JOIN         0x00004000 /* Received Join */
 533 #define TMPF_RESET              0x00008000 /* Send RST */
 534 #define TMPF_TCP_FALLBACK       0x00010000 /* Fallback to TCP */
 535 #define TMPF_FASTCLOSERCV       0x00020000 /* Received Fastclose option */
 536 #define TMPF_EMBED_DSN          0x00040000 /* tp has DSN mapping */
 537 #define TMPF_MPTCP_READY        0x00080000 /* Can send DSS options on data */
 538 #define TMPF_INFIN_SENT         0x00100000 /* Sent infinite mapping */
 539 #define TMPF_SND_MPFAIL         0x00200000 /* Received mapping csum failure */
 540 #define TMPF_FASTJOIN_SEND      0x00400000 /* Fast join early data send */
 541 #define TMPF_FASTJOINBY2_SEND   0x00800000 /* Fast join send after 3 WHS */
 542 #define TMPF_TFO_REQUEST        0x02000000 /* TFO Requested */
 543
 544         tcp_seq                 t_mpuna;        /* unacknowledged sequence */
 545         void                    *t_mptcb;       /* pointer to MPTCP TCB */
 546         struct mpt_dsn_map      t_rcv_map;      /* Receive mapping list */
 547         u_int8_t                t_local_aid;    /* Addr Id for authentication */
 548         u_int8_t                t_rem_aid;      /* Addr ID of another subflow */
 549         u_int8_t                t_mprxtshift;   /* join retransmission */
 550 #endif /* MPTCP */
 551
 552 #define TFO_F_OFFER_COOKIE      0x01 /* We will offer a cookie */
 553 #define TFO_F_COOKIE_VALID      0x02 /* The received cookie is valid */
 554 #define TFO_F_COOKIE_REQ        0x04 /* Client requested a new cookie */
 555 #define TFO_F_COOKIE_SENT       0x08 /* Client did send a cookie in the SYN */
 556 #define TFO_F_SYN_LOSS          0x10 /* A SYN-loss triggered a fallback to regular TCP on the client-side */
 557 #define TFO_F_NO_RCVPROBING     0x20 /* This network is guaranteed to support TFO in the downstream direction */
 558 #define TFO_F_NO_SNDPROBING     0x40 /* This network is guaranteed to support TFO in the upstream direction */
 559         u_int8_t                t_tfo_flags;
 560 #define TFO_S_SYNDATA_RCV       0x01 /* SYN+data has been received */
 561 #define TFO_S_COOKIEREQ_RECV    0x02 /* TFO-cookie request received */
 562 #define TFO_S_COOKIE_SENT       0x04 /* TFO-cookie announced in SYN/ACK */
 563 #define TFO_S_COOKIE_INVALID    0x08 /* Received TFO-cookie is invalid */
 564 #define TFO_S_COOKIE_REQ        0x10 /* TFO-cookie requested within the SYN */
 565 #define TFO_S_COOKIE_RCV        0x20 /* TFO-cookie received in SYN/ACK */
 566 #define TFO_S_SYN_DATA_SENT     0x40 /* SYN+data sent */
 567 #define TFO_S_SYN_DATA_ACKED    0x80 /* SYN+data has been acknowledged in SYN/ACK */
 568 #define TFO_S_SYN_LOSS          0x0100 /* SYN+TFO has been lost - fallback to regular TCP */
 569 #define TFO_S_COOKIE_WRONG      0x0200 /* Cookie we sent in the SYN was wrong */
 570 #define TFO_S_NO_COOKIE_RCV     0x0400 /* We asked for a cookie but didn't get one */
 571 #define TFO_S_HEURISTICS_DISABLE 0x0800 /* TFO-heuristics disabled it for this connection */
 572 #define TFO_S_SEND_BLACKHOLE    0x1000 /* TFO got blackholed in the send direction */
 573 #define TFO_S_RECV_BLACKHOLE    0x2000 /* TFO got blackholed in the recv direction */
 574         u_int16_t               t_tfo_stats;
 575
 576         u_int8_t                t_tfo_probes; /* TFO-probes we did send */
 577 /*
 578  * This here is the TFO-probing state-machine. Transitions are as follows:
 579  *
 580  * Current state: PROBE_NONE
 581  *                Event: SYN+DATA acknowledged
 582  *                       Action: Transition to PROBE_PROBING and set keepalive-timer
 583  *
 584  * Current state: PROBE_PROBING (initial state)
 585  *                Event: Receive data
 586  *                       Action: Transition to PROBE_NONE and cancel keepalive-timer
 587  *                Event: Receive ACK that does not indicate a hole
 588  *                       Action: Transition to PROBE_NONE and cancel keepalive-timer
 589  *                Event: Receive ACK that indicates a hole
 590  *                       Action: Transition to PROBE_WAIT_DATA and set a short timer
 591  *                               to wait for the final segment.
 592  *                Event: Keepalive-timeout (did not receive any segment)
 593  *                       Action: Signal ETIMEDOUT as with regular keepalive-timers
 594  *
 595  * Current state: PROBE_WAIT_DATA
 596  *                Event: Receive data
 597  *                       Action: Transition to PROBE_NONE and cancel keepalive-timer
 598  *                Event: Data-timeout (did not receive the expected data)
 599  *                       Action: Signal ENODATA up to the app and close everything.
 600  */
 601 #define TFO_PROBE_NONE          0 /* Not probing now */
 602 #define TFO_PROBE_PROBING       1 /* Sending out TCP-keepalives waiting for reply */
 603 #define TFO_PROBE_WAIT_DATA     2 /* Received reply, waiting for data */
 604         u_int8_t                t_tfo_probe_state;
 605
 606         u_int32_t       t_rcvoopack;            /* out-of-order packets received */
 607         u_int32_t       t_pawsdrop;             /* segments dropped due to PAWS */
 608         u_int32_t       t_sack_recovery_episode; /* SACK recovery episodes */
 609         u_int32_t       t_reordered_pkts;       /* packets reorderd */
 610         u_int32_t       t_dsack_sent;           /* Sent DSACK notification */
 611         u_int32_t       t_dsack_recvd;          /* Received a valid DSACK option */
 612         SLIST_HEAD(,tcp_notify_ack_marker) t_notify_ack; /* state for notifying data acknowledgements */
 613         u_int32_t       t_recv_throttle_ts;     /* TS for start of recv throttle */
 614 };
 615
 616 #define IN_FASTRECOVERY(tp)     (tp->t_flags & TF_FASTRECOVERY)
 617 #define SACK_ENABLED(tp)        (tp->t_flagsext & TF_SACK_ENABLE)
 618
 619 /*
 620  * If the connection is in a throttled state due to advisory feedback from
 621  * the interface output queue, reset that state. We do this in favor
 622  * of entering recovery because the data transfer during recovery
 623  * should be just a trickle and it will help to improve performance.
 624  * We also do not want to back off twice in the same RTT.
 625  */
 626 #define ENTER_FASTRECOVERY(_tp_) do {                           \
 627         (_tp_)->t_flags |= TF_FASTRECOVERY;                     \
 628         if (INP_IS_FLOW_CONTROLLED((_tp_)->t_inpcb))            \
 629                 inp_reset_fc_state((_tp_)->t_inpcb);            \
 630         if (!SLIST_EMPTY(&tp->t_rxt_segments))                  \
 631                 tcp_rxtseg_clean(tp);                           \
 632 } while(0)
 633
 634 #define EXIT_FASTRECOVERY(_tp_) do {            \
 635         (_tp_)->t_flags &= ~TF_FASTRECOVERY;    \
 636         (_tp_)->t_dupacks = 0;                  \
 637         (_tp_)->t_rexmtthresh = tcprexmtthresh; \
 638         (_tp_)->t_bytes_acked = 0;              \
 639         (_tp_)->ecn_flags &= ~TE_INRECOVERY;    \
 640         (_tp_)->t_timer[TCPT_PTO] = 0;          \
 641         (_tp_)->t_flagsext &= ~TF_RESCUE_RXT;   \
 642         (_tp_)->t_lossflightsize = 0;           \
 643 } while(0)
 644
 645 /*
 646  * When the number of duplicate acks received is less than
 647  * the retransmit threshold, use Limited Transmit algorithm
 648  */
 649 extern int tcprexmtthresh;
 650 #define ALLOW_LIMITED_TRANSMIT(_tp_) \
 651         ((_tp_)->t_dupacks > 0 && \
 652         (_tp_)->t_dupacks < (_tp_)->t_rexmtthresh && \
 653         ((_tp_)->t_flagsext & (TF_PKTS_REORDERED|TF_DELAY_RECOVERY)) \
 654             != (TF_PKTS_REORDERED|TF_DELAY_RECOVERY))
 655
 656 /*
 657  * This condition is true is timestamp option is supported
 658  * on a connection.
 659  */
 660 #define TSTMP_SUPPORTED(_tp_) \
 661         (((_tp_)->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == \
 662                 (TF_REQ_TSTMP|TF_RCVD_TSTMP))
 663
 664 /*
 665  * This condition is true if window scale option is supported
 666  * on a connection
 667  */
 668 #define TCP_WINDOW_SCALE_ENABLED(_tp_) \
 669         (((_tp_)->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == \
 670                 (TF_RCVD_SCALE|TF_REQ_SCALE))
 671
 672 /* Is ECN enabled end-to-end */
 673 #define TCP_ECN_ENABLED(_tp_) \
 674         (((_tp_)->ecn_flags & (TE_ECN_ON)) == (TE_ECN_ON))
 675
 676 /*
 677  * Gives number of bytes acked by this ack
 678  */
 679 #define BYTES_ACKED(_th_, _tp_) \
 680         ((_th_)->th_ack - (_tp_)->snd_una)
 681
 682 /* Returns true if a DSACK option should be sent */
 683 #define TCP_SEND_DSACK_OPT(_tp_) \
 684         ((_tp_)->t_dsack_lseq > 0 && (_tp_)->t_dsack_rseq > 0)
 685
 686 /* Check if DSACK option should be processed */
 687 #define TCP_DSACK_ENABLED(tp)   (tcp_dsack_enable == 1 && \
 688     !(tp->t_flagsext & TF_DISABLE_DSACK))
 689
 690 /*
 691  * Returns true if a DSACK sequence is within the max send window that will
 692  * be accepted. In order to set a window to validate sequence numbers, the
 693  * max send window within which a DSACK option is processed is limited.
 694  *
 695  * We need to choose a maximum window to check if the sequence number is
 696  * within the window. One arbitrary choice is 256 * MSS because if the
 697  * window is as large as 256 segments it might be big enough to ignore the
 698  * DSACK option. Choosing a much larger limit means that the memory for
 699  * retransmit segments can be held for a longer time.
 700  */
 701 #define TCP_DSACK_MAX_SEND_WINDOW(_tp_) ((_tp_)->t_maxseg << 8)
 702 #define TCP_DSACK_SEQ_IN_WINDOW(_tp_, _seq_, _una_) \
 703     (SEQ_LEQ((_seq_), (_tp_)->snd_max) && \
 704     SEQ_GEQ((_seq_), ((_una_) - TCP_DSACK_MAX_SEND_WINDOW(_tp_))))
 705
 706 #define TCP_RESET_REXMT_STATE(_tp_) do { \
 707         (_tp_)->t_rxtshift = 0; \
 708         (_tp_)->t_rxtstart = 0; \
 709 } while(0);
 710
 711 #define TCP_AUTORCVBUF_MAX(_ifp_) (((_ifp_) != NULL && \
 712         ((_ifp_)->if_eflags & IFEF_3CA)) ? tcp_autorcvbuf_max_ca : \
 713         tcp_autorcvbuf_max)
 714
 715 enum tcp_cc_event {
 716         TCP_CC_CWND_INIT,       /* 0 */
 717         TCP_CC_INSEQ_ACK_RCVD,  /* 1 */
 718         TCP_CC_ACK_RCVD,        /* 2 */
 719         TCP_CC_ENTER_FASTRECOVERY, /* 3 */
 720         TCP_CC_IN_FASTRECOVERY, /* 4 */
 721         TCP_CC_EXIT_FASTRECOVERY,  /* 5 */
 722         TCP_CC_PARTIAL_ACK,     /* 6 */
 723         TCP_CC_IDLE_TIMEOUT,    /* 7 */
 724         TCP_CC_REXMT_TIMEOUT,   /* 8 */
 725         TCP_CC_ECN_RCVD,        /* 9 */
 726         TCP_CC_BAD_REXMT_RECOVERY, /* 10 */
 727         TCP_CC_OUTPUT_ERROR,    /* 11 */
 728         TCP_CC_CHANGE_ALGO,     /* 12 */
 729         TCP_CC_FLOW_CONTROL,    /* 13 */
 730         TCP_CC_SUSPEND,         /* 14 */
 731         TCP_CC_LIMITED_TRANSMIT, /* 15 */
 732         TCP_CC_EARLY_RETRANSMIT, /* 16 */
 733         TCP_CC_TLP_RECOVERY,    /* 17 */
 734         TCP_CC_TLP_RECOVER_LASTPACKET, /* 18 */
 735         TCP_CC_DELAY_FASTRECOVERY, /* 19 */
 736         TCP_CC_TLP_IN_FASTRECOVERY, /* 20 */
 737         TCP_CC_DSACK_BAD_REXMT  /* 21 */
 738 };
 739
 740 /*
 741  * Structure to hold TCP options that are only used during segment
 742  * processing (in tcp_input), but not held in the tcpcb.
 743  * It's basically used to reduce the number of parameters
 744  * to tcp_dooptions.
 745  */
 746 struct tcpopt {
 747         u_int32_t       to_flags;               /* which options are present */
 748 #define TOF_TS          0x0001          /* timestamp */
 749 #define TOF_MSS         0x0010
 750 #define TOF_SCALE       0x0020
 751 #define TOF_SIGNATURE   0x0040  /* signature option present */
 752 #define TOF_SIGLEN      0x0080  /* signature length valid (RFC2385) */
 753 #define TOF_SACK        0x0100          /* Peer sent SACK option */
 754 #define TOF_MPTCP       0x0200  /* MPTCP options to be dropped */
 755 #define TOF_TFO         0x0400  /* TFO cookie option present */
 756 #define TOF_TFOREQ      0x0800  /* TFO cookie request present */
 757         u_int32_t               to_tsval;
 758         u_int32_t               to_tsecr;
 759         u_int16_t       to_mss;
 760         u_int8_t        to_requested_s_scale;
 761         u_int8_t        to_nsacks;      /* number of SACK blocks */
 762         u_char          *to_sacks;      /* pointer to the first SACK blocks */
 763         u_char          *to_tfo;        /* pointer to the TFO cookie */
 764 };
 765
 766 /*
 767  * The TAO cache entry which is stored in the protocol family specific
 768  * portion of the route metrics.
 769  */
 770 struct rmxp_tao {
 771         tcp_cc  tao_cc;                 /* latest CC in valid SYN */
 772         tcp_cc  tao_ccsent;             /* latest CC sent to peer */
 773         u_short tao_mssopt;             /* peer's cached MSS */
 774 #ifdef notyet
 775         u_short tao_flags;              /* cache status flags */
 776 #define TAOF_DONT       0x0001          /* peer doesn't understand rfc1644 */
 777 #define TAOF_OK         0x0002          /* peer does understand rfc1644 */
 778 #define TAOF_UNDEF      0               /* we don't know yet */
 779 #endif /* notyet */
 780 };
 781 #define rmx_taop(r)     ((struct rmxp_tao *)(r).rmx_filler)
 782
 783 #define intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
 784 #define sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
 785
 786 /*
 787  * The rtt measured is in milliseconds as the timestamp granularity is
 788  * a millisecond. The smoothed round-trip time and estimated variance
 789  * are stored as fixed point numbers scaled by the values below.
 790  * For convenience, these scales are also used in smoothing the average
 791  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
 792  * With these scales, srtt has 5 bits to the right of the binary point,
 793  * and thus an "ALPHA" of 0.875.  rttvar has 4 bits to the right of the
 794  * binary point, and is smoothed with an ALPHA of 0.75.
 795  */
 796 #define TCP_RTT_SCALE           32      /* multiplier for srtt; 3 bits frac. */
 797 #define TCP_RTT_SHIFT           5       /* shift for srtt; 5 bits frac. */
 798 #define TCP_RTTVAR_SCALE        16      /* multiplier for rttvar; 4 bits */
 799 #define TCP_RTTVAR_SHIFT        4       /* shift for rttvar; 4 bits */
 800 #define TCP_DELTA_SHIFT         2       /* see tcp_input.c */
 801
 802 /* TFO-specific defines */
 803 #define TFO_COOKIE_LEN_MIN      4
 804 #define TFO_COOKIE_LEN_DEFAULT  8
 805 #define TFO_COOKIE_LEN_MAX      16
 806
 807 /*
 808  * The initial retransmission should happen at rtt + 4 * rttvar.
 809  * Because of the way we do the smoothing, srtt and rttvar
 810  * will each average +1/2 tick of bias.  When we compute
 811  * the retransmit timer, we want 1/2 tick of rounding and
 812  * 1 extra tick because of +-1/2 tick uncertainty in the
 813  * firing of the timer.  The bias will give us exactly the
 814  * 1.5 tick we need.  But, because the bias is
 815  * statistical, we have to test that we don't drop below
 816  * the minimum feasible timer (which is 2 ticks).
 817  * This version of the macro adapted from a paper by Lawrence
 818  * Brakmo and Larry Peterson which outlines a problem caused
 819  * by insufficient precision in the original implementation,
 820  * which results in inappropriately large RTO values for very
 821  * fast networks.
 822  */
 823 #define TCP_REXMTVAL(tp) \
 824         max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
 825           + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
 826
 827 /*
 828  * Jaguar compatible TCP control block, for xtcpcb
 829  * Does not have the old fields
 830  */
 831 struct otcpcb {
 832 #else
 833 struct tseg_qent;
 834 _TCPCB_LIST_HEAD(tsegqe_head, tseg_qent);
 835
 836 struct tcpcb {
 837 #endif /* KERNEL_PRIVATE */
 838 #if defined(KERNEL_PRIVATE)
 839         u_int32_t t_segq;
 840 #else
 841         struct  tsegqe_head t_segq;
 842 #endif /* KERNEL_PRIVATE */
 843         int     t_dupacks;              /* consecutive dup acks recd */
 844         u_int32_t unused;               /* unused now: was t_template */
 845
 846         int     t_timer[TCPT_NTIMERS_EXT];      /* tcp timers */
 847
 848         _TCPCB_PTR(struct inpcb *) t_inpcb;     /* back pointer to internet pcb */
 849         int     t_state;                /* state of this connection */
 850         u_int   t_flags;
 851 #define TF_ACKNOW       0x00001         /* ack peer immediately */
 852 #define TF_DELACK       0x00002         /* ack, but try to delay it */
 853 #define TF_NODELAY      0x00004         /* don't delay packets to coalesce */
 854 #define TF_NOOPT        0x00008         /* don't use tcp options */
 855 #define TF_SENTFIN      0x00010         /* have sent FIN */
 856 #define TF_REQ_SCALE    0x00020         /* have/will request window scaling */
 857 #define TF_RCVD_SCALE   0x00040         /* other side has requested scaling */
 858 #define TF_REQ_TSTMP    0x00080         /* have/will request timestamps */
 859 #define TF_RCVD_TSTMP   0x00100         /* a timestamp was received in SYN */
 860 #define TF_SACK_PERMIT  0x00200         /* other side said I could SACK */
 861 #define TF_NEEDSYN      0x00400         /* send SYN (implicit state) */
 862 #define TF_NEEDFIN      0x00800         /* send FIN (implicit state) */
 863 #define TF_NOPUSH       0x01000         /* don't push */
 864 #define TF_REQ_CC       0x02000         /* have/will request CC */
 865 #define TF_RCVD_CC      0x04000         /* a CC was received in SYN */
 866 #define TF_SENDCCNEW    0x08000         /* Not implemented */
 867 #define TF_MORETOCOME   0x10000         /* More data to be appended to sock */
 868 #define TF_LQ_OVERFLOW  0x20000         /* listen queue overflow */
 869 #define TF_RXWIN0SENT   0x40000         /* sent a receiver win 0 in response */
 870 #define TF_SLOWLINK     0x80000         /* route is a on a modem speed link */
 871
 872         int     t_force;                /* 1 if forcing out a byte */
 873
 874         tcp_seq snd_una;                /* send unacknowledged */
 875         tcp_seq snd_max;                /* highest sequence number sent;
 876                                          * used to recognize retransmits
 877                                          */
 878         tcp_seq snd_nxt;                /* send next */
 879         tcp_seq snd_up;                 /* send urgent pointer */
 880
 881         tcp_seq snd_wl1;                /* window update seg seq number */
 882         tcp_seq snd_wl2;                /* window update seg ack number */
 883         tcp_seq iss;                    /* initial send sequence number */
 884         tcp_seq irs;                    /* initial receive sequence number */
 885
 886         tcp_seq rcv_nxt;                /* receive next */
 887         tcp_seq rcv_adv;                /* advertised window */
 888         u_int32_t rcv_wnd;              /* receive window */
 889         tcp_seq rcv_up;                 /* receive urgent pointer */
 890
 891         u_int32_t snd_wnd;              /* send window */
 892         u_int32_t snd_cwnd;             /* congestion-controlled window */
 893         u_int32_t snd_ssthresh;         /* snd_cwnd size threshold for
 894                                          * for slow start exponential to
 895                                          * linear switch
 896                                          */
 897         u_int   t_maxopd;               /* mss plus options */
 898
 899         u_int32_t t_rcvtime;            /* time at which a packet was received */
 900         u_int32_t t_starttime;          /* time connection was established */
 901         int     t_rtttime;              /* round trip time */
 902         tcp_seq t_rtseq;                /* sequence number being timed */
 903
 904         int     t_rxtcur;               /* current retransmit value (ticks) */
 905         u_int   t_maxseg;               /* maximum segment size */
 906         int     t_srtt;                 /* smoothed round-trip time */
 907         int     t_rttvar;               /* variance in round-trip time */
 908
 909         int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
 910         u_int   t_rttmin;               /* minimum rtt allowed */
 911         u_int32_t t_rttupdated;         /* number of times rtt sampled */
 912         u_int32_t max_sndwnd;           /* largest window peer has offered */
 913
 914         int     t_softerror;            /* possible error not yet reported */
 915 /* out-of-band data */
 916         char    t_oobflags;             /* have some */
 917         char    t_iobc;                 /* input character */
 918 #define TCPOOB_HAVEDATA 0x01
 919 #define TCPOOB_HADDATA  0x02
 920 /* RFC 1323 variables */
 921         u_char  snd_scale;              /* window scaling for send window */
 922         u_char  rcv_scale;              /* window scaling for recv window */
 923         u_char  request_r_scale;        /* pending window scaling */
 924         u_char  requested_s_scale;
 925         u_int32_t ts_recent;            /* timestamp echo data */
 926
 927         u_int32_t ts_recent_age;        /* when last updated */
 928         tcp_seq last_ack_sent;
 929 /* RFC 1644 variables */
 930         tcp_cc  cc_send;                /* send connection count */
 931         tcp_cc  cc_recv;                /* receive connection count */
 932         tcp_seq snd_recover;            /* for use in fast recovery */
 933 /* experimental */
 934         u_int32_t snd_cwnd_prev;        /* cwnd prior to retransmit */
 935         u_int32_t snd_ssthresh_prev;    /* ssthresh prior to retransmit */
 936         u_int32_t t_badrxtwin;          /* window for retransmit recovery */
 937 };
 938
 939 #define        tcps_ecn_setup  tcps_ecn_client_success
 940 #define        tcps_sent_cwr   tcps_ecn_recv_ece
 941 #define        tcps_sent_ece   tcps_ecn_sent_ece
 942
 943 /*
 944  * TCP statistics.
 945  * Many of these should be kept per connection,
 946  * but that's inconvenient at the moment.
 947  */
 948 struct  tcpstat {
 949         u_int32_t       tcps_connattempt;       /* connections initiated */
 950         u_int32_t       tcps_accepts;           /* connections accepted */
 951         u_int32_t       tcps_connects;          /* connections established */
 952         u_int32_t       tcps_drops;             /* connections dropped */
 953         u_int32_t       tcps_conndrops;         /* embryonic connections dropped */
 954         u_int32_t       tcps_closed;            /* conn. closed (includes drops) */
 955         u_int32_t       tcps_segstimed;         /* segs where we tried to get rtt */
 956         u_int32_t       tcps_rttupdated;        /* times we succeeded */
 957         u_int32_t       tcps_delack;            /* delayed acks sent */
 958         u_int32_t       tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
 959         u_int32_t       tcps_rexmttimeo;        /* retransmit timeouts */
 960         u_int32_t       tcps_persisttimeo;      /* persist timeouts */
 961         u_int32_t       tcps_keeptimeo;         /* keepalive timeouts */
 962         u_int32_t       tcps_keepprobe;         /* keepalive probes sent */
 963         u_int32_t       tcps_keepdrops;         /* connections dropped in keepalive */
 964
 965         u_int32_t       tcps_sndtotal;          /* total packets sent */
 966         u_int32_t       tcps_sndpack;           /* data packets sent */
 967         u_int32_t       tcps_sndbyte;           /* data bytes sent */
 968         u_int32_t       tcps_sndrexmitpack;     /* data packets retransmitted */
 969         u_int32_t       tcps_sndrexmitbyte;     /* data bytes retransmitted */
 970         u_int32_t       tcps_sndacks;           /* ack-only packets sent */
 971         u_int32_t       tcps_sndprobe;          /* window probes sent */
 972         u_int32_t       tcps_sndurg;            /* packets sent with URG only */
 973         u_int32_t       tcps_sndwinup;          /* window update-only packets sent */
 974         u_int32_t       tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
 975
 976         u_int32_t       tcps_rcvtotal;          /* total packets received */
 977         u_int32_t       tcps_rcvpack;           /* packets received in sequence */
 978         u_int32_t       tcps_rcvbyte;           /* bytes received in sequence */
 979         u_int32_t       tcps_rcvbadsum;         /* packets received with ccksum errs */
 980         u_int32_t       tcps_rcvbadoff;         /* packets received with bad offset */
 981         u_int32_t       tcps_rcvmemdrop;        /* packets dropped for lack of memory */
 982         u_int32_t       tcps_rcvshort;          /* packets received too short */
 983         u_int32_t       tcps_rcvduppack;        /* duplicate-only packets received */
 984         u_int32_t       tcps_rcvdupbyte;        /* duplicate-only bytes received */
 985         u_int32_t       tcps_rcvpartduppack;    /* packets with some duplicate data */
 986         u_int32_t       tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
 987         u_int32_t       tcps_rcvoopack;         /* out-of-order packets received */
 988         u_int32_t       tcps_rcvoobyte;         /* out-of-order bytes received */
 989         u_int32_t       tcps_rcvpackafterwin;   /* packets with data after window */
 990         u_int32_t       tcps_rcvbyteafterwin;   /* bytes rcvd after window */
 991         u_int32_t       tcps_rcvafterclose;     /* packets rcvd after "close" */
 992         u_int32_t       tcps_rcvwinprobe;       /* rcvd window probe packets */
 993         u_int32_t       tcps_rcvdupack;         /* rcvd duplicate acks */
 994         u_int32_t       tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
 995         u_int32_t       tcps_rcvackpack;        /* rcvd ack packets */
 996         u_int32_t       tcps_rcvackbyte;        /* bytes acked by rcvd acks */
 997         u_int32_t       tcps_rcvwinupd;         /* rcvd window update packets */
 998         u_int32_t       tcps_pawsdrop;          /* segments dropped due to PAWS */
 999         u_int32_t       tcps_predack;           /* times hdr predict ok for acks */
1000         u_int32_t       tcps_preddat;           /* times hdr predict ok for data pkts */
1001         u_int32_t       tcps_pcbcachemiss;
1002         u_int32_t       tcps_cachedrtt;         /* times cached RTT in route updated */
1003         u_int32_t       tcps_cachedrttvar;      /* times cached rttvar updated */
1004         u_int32_t       tcps_cachedssthresh;    /* times cached ssthresh updated */
1005         u_int32_t       tcps_usedrtt;           /* times RTT initialized from route */
1006         u_int32_t       tcps_usedrttvar;        /* times RTTVAR initialized from rt */
1007         u_int32_t       tcps_usedssthresh;      /* times ssthresh initialized from rt*/
1008         u_int32_t       tcps_persistdrop;       /* timeout in persist state */
1009         u_int32_t       tcps_badsyn;            /* bogus SYN, e.g. premature ACK */
1010         u_int32_t       tcps_mturesent;         /* resends due to MTU discovery */
1011         u_int32_t       tcps_listendrop;        /* listen queue overflows */
1012
1013         /* new stats from FreeBSD 5.4 sync up */
1014         u_int32_t       tcps_minmssdrops;       /* average minmss too low drops */
1015         u_int32_t       tcps_sndrexmitbad;      /* unnecessary packet retransmissions */
1016         u_int32_t       tcps_badrst;            /* ignored RSTs in the window */
1017
1018         u_int32_t       tcps_sc_added;          /* entry added to syncache */
1019         u_int32_t       tcps_sc_retransmitted;  /* syncache entry was retransmitted */
1020         u_int32_t       tcps_sc_dupsyn;         /* duplicate SYN packet */
1021         u_int32_t       tcps_sc_dropped;        /* could not reply to packet */
1022         u_int32_t       tcps_sc_completed;      /* successful extraction of entry */
1023         u_int32_t       tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
1024         u_int32_t       tcps_sc_cacheoverflow;  /* syncache cache limit hit */
1025         u_int32_t       tcps_sc_reset;          /* RST removed entry from syncache */
1026         u_int32_t       tcps_sc_stale;          /* timed out or listen socket gone */
1027         u_int32_t       tcps_sc_aborted;        /* syncache entry aborted */
1028         u_int32_t       tcps_sc_badack;         /* removed due to bad ACK */
1029         u_int32_t       tcps_sc_unreach;        /* ICMP unreachable received */
1030         u_int32_t       tcps_sc_zonefail;       /* zalloc() failed */
1031         u_int32_t       tcps_sc_sendcookie;     /* SYN cookie sent */
1032         u_int32_t       tcps_sc_recvcookie;     /* SYN cookie received */
1033
1034         u_int32_t       tcps_hc_added;          /* entry added to hostcache */
1035         u_int32_t       tcps_hc_bucketoverflow; /* hostcache per bucket limit hit */
1036
1037         /* SACK related stats */
1038         u_int32_t       tcps_sack_recovery_episode; /* SACK recovery episodes */
1039         u_int32_t       tcps_sack_rexmits;          /* SACK rexmit segments   */
1040         u_int32_t       tcps_sack_rexmit_bytes;     /* SACK rexmit bytes      */
1041         u_int32_t       tcps_sack_rcv_blocks;       /* SACK blocks (options) received */
1042         u_int32_t       tcps_sack_send_blocks;      /* SACK blocks (options) sent     */
1043         u_int32_t       tcps_sack_sboverflow;       /* SACK sendblock overflow   */
1044
1045         u_int32_t       tcps_bg_rcvtotal;       /* total background packets received */
1046         u_int32_t       tcps_rxtfindrop;        /* drop conn after retransmitting FIN */
1047         u_int32_t       tcps_fcholdpacket;      /* packets withheld because of flow control */
1048
1049         /* LRO related stats */
1050         u_int32_t       tcps_coalesced_pack;    /* number of coalesced packets */
1051         u_int32_t       tcps_flowtbl_full;      /* times flow table was full */
1052         u_int32_t       tcps_flowtbl_collision; /* collisions in flow tbl */
1053         u_int32_t       tcps_lro_twopack;       /* 2 packets coalesced */
1054         u_int32_t       tcps_lro_multpack;      /* 3 or 4 pkts coalesced */
1055         u_int32_t       tcps_lro_largepack;     /* 5 or more pkts coalesced */
1056
1057         u_int32_t       tcps_limited_txt;       /* Limited transmit used */
1058         u_int32_t       tcps_early_rexmt;       /* Early retransmit used */
1059         u_int32_t       tcps_sack_ackadv;       /* Cumulative ack advanced along with sack */
1060
1061         /* Checksum related stats */
1062         u_int32_t       tcps_rcv_swcsum;        /* tcp swcksum (inbound), packets */
1063         u_int32_t       tcps_rcv_swcsum_bytes;  /* tcp swcksum (inbound), bytes */
1064         u_int32_t       tcps_rcv6_swcsum;       /* tcp6 swcksum (inbound), packets */
1065         u_int32_t       tcps_rcv6_swcsum_bytes; /* tcp6 swcksum (inbound), bytes */
1066         u_int32_t       tcps_snd_swcsum;        /* tcp swcksum (outbound), packets */
1067         u_int32_t       tcps_snd_swcsum_bytes;  /* tcp swcksum (outbound), bytes */
1068         u_int32_t       tcps_snd6_swcsum;       /* tcp6 swcksum (outbound), packets */
1069         u_int32_t       tcps_snd6_swcsum_bytes; /* tcp6 swcksum (outbound), bytes */
1070         u_int32_t       tcps_msg_unopkts;       /* unordered packet on TCP msg stream */
1071         u_int32_t       tcps_msg_unoappendfail; /* failed to append unordered pkt */
1072         u_int32_t       tcps_msg_sndwaithipri; /* send is waiting for high priority data */
1073
1074         /* MPTCP Related stats */
1075         u_int32_t       tcps_invalid_mpcap;     /* Invalid MPTCP capable opts */
1076         u_int32_t       tcps_invalid_joins;     /* Invalid MPTCP joins */
1077         u_int32_t       tcps_mpcap_fallback;    /* TCP fallback in primary */
1078         u_int32_t       tcps_join_fallback;     /* No MPTCP in secondary */
1079         u_int32_t       tcps_estab_fallback;    /* DSS option dropped */
1080         u_int32_t       tcps_invalid_opt;       /* Catchall error stat */
1081         u_int32_t       tcps_mp_outofwin;       /* Packet lies outside the
1082                                                    shared recv window */
1083         u_int32_t       tcps_mp_reducedwin;     /* Reduced subflow window */
1084         u_int32_t       tcps_mp_badcsum;        /* Bad DSS csum */
1085         u_int32_t       tcps_mp_oodata;         /* Out of order data */
1086         u_int32_t       tcps_mp_switches;       /* number of subflow switch */
1087         u_int32_t       tcps_mp_rcvtotal;       /* number of rcvd packets */
1088         u_int32_t       tcps_mp_rcvbytes;       /* number of bytes received */
1089         u_int32_t       tcps_mp_sndpacks;       /* number of data packs sent */
1090         u_int32_t       tcps_mp_sndbytes;       /* number of bytes sent */
1091         u_int32_t       tcps_join_rxmts;        /* join ack retransmits */
1092         u_int32_t       tcps_tailloss_rto;      /* RTO due to tail loss */
1093         u_int32_t       tcps_reordered_pkts;    /* packets reorderd */
1094         u_int32_t       tcps_recovered_pkts;    /* recovered after loss */
1095         u_int32_t       tcps_pto;               /* probe timeout */
1096         u_int32_t       tcps_rto_after_pto;     /* RTO after a probe */
1097         u_int32_t       tcps_tlp_recovery;      /* TLP induced fast recovery */
1098         u_int32_t       tcps_tlp_recoverlastpkt; /* TLP recoverd last pkt */
1099         u_int32_t       tcps_ecn_client_success; /* client-side connection negotiated ECN */
1100         u_int32_t       tcps_ecn_recv_ece;      /* ECE received, sent CWR */
1101         u_int32_t       tcps_ecn_sent_ece;      /* Sent ECE notification */
1102         u_int32_t       tcps_detect_reordering; /* Detect pkt reordering */
1103         u_int32_t       tcps_delay_recovery;    /* Delay fast recovery */
1104         u_int32_t       tcps_avoid_rxmt;        /* Retransmission was avoided */
1105         u_int32_t       tcps_unnecessary_rxmt;  /* Retransmission was not needed */
1106         u_int32_t       tcps_nostretchack;      /* disabled stretch ack algorithm on a connection */
1107         u_int32_t       tcps_rescue_rxmt;       /* SACK rescue retransmit */
1108         u_int32_t       tcps_pto_in_recovery;   /* rescue retransmit in fast recovery */
1109         u_int32_t       tcps_pmtudbh_reverted;  /* PMTU Blackhole detection, segment size reverted */
1110
1111         /* DSACK related statistics */
1112         u_int32_t       tcps_dsack_disable;     /* DSACK disabled due to n/w duplication */
1113         u_int32_t       tcps_dsack_ackloss;     /* ignore DSACK due to ack loss */
1114         u_int32_t       tcps_dsack_badrexmt;    /* DSACK based bad rexmt recovery */
1115         u_int32_t       tcps_dsack_sent;        /* Sent DSACK notification */
1116         u_int32_t       tcps_dsack_recvd;       /* Received a valid DSACK option */
1117         u_int32_t       tcps_dsack_recvd_old;   /* Received an out of window DSACK option */
1118
1119         /* MPTCP Subflow selection stats */
1120         u_int32_t       tcps_mp_sel_symtomsd;   /* By symptomsd */
1121         u_int32_t       tcps_mp_sel_rtt;        /* By RTT comparison */
1122         u_int32_t       tcps_mp_sel_rto;        /* By RTO comparison */
1123         u_int32_t       tcps_mp_sel_peer;       /* By peer's output pattern */
1124         u_int32_t       tcps_mp_num_probes;     /* Number of probes sent */
1125         u_int32_t       tcps_mp_verdowngrade;   /* MPTCP version downgrade */
1126         u_int32_t       tcps_drop_after_sleep;  /* drop after long AP sleep */
1127         u_int32_t       tcps_probe_if;          /* probe packets after interface availability */
1128         u_int32_t       tcps_probe_if_conflict; /* Can't send probe packets for interface */
1129
1130         u_int32_t       tcps_ecn_client_setup;  /* Attempted ECN setup from client side */
1131         u_int32_t       tcps_ecn_server_setup;  /* Attempted ECN setup from server side */
1132         u_int32_t       tcps_ecn_server_success; /* server-side connection negotiated ECN */
1133         u_int32_t       tcps_ecn_lost_synack;   /* Lost SYN-ACK with ECN setup */
1134         u_int32_t       tcps_ecn_lost_syn;      /* Lost SYN with ECN setup */
1135         u_int32_t       tcps_ecn_not_supported; /* Server did not support ECN setup */
1136         u_int32_t       tcps_ecn_recv_ce;       /* Received CE from the network */
1137         u_int32_t       tcps_ecn_conn_recv_ce;  /* Number of connections received CE atleast once */
1138         u_int32_t       tcps_ecn_conn_recv_ece; /* Number of connections received ECE atleast once */
1139         u_int32_t       tcps_ecn_conn_plnoce;   /* Number of connections that received no CE and sufferred packet loss */
1140         u_int32_t       tcps_ecn_conn_pl_ce;    /* Number of connections that received CE and sufferred packet loss */
1141         u_int32_t       tcps_ecn_conn_nopl_ce;  /* Number of connections that received CE and sufferred no packet loss */
1142         u_int32_t       tcps_ecn_fallback_synloss; /* Number of times we did fall back due to SYN-Loss */
1143         u_int32_t       tcps_ecn_fallback_reorder; /* Number of times we fallback because we detected the PAWS-issue */
1144         u_int32_t       tcps_ecn_fallback_ce;   /* Number of times we fallback because we received too many CEs */
1145
1146         /* TFO-related statistics */
1147         u_int32_t       tcps_tfo_syn_data_rcv;  /* Received a SYN+data with valid cookie */
1148         u_int32_t       tcps_tfo_cookie_req_rcv;/* Received a TFO cookie-request */
1149         u_int32_t       tcps_tfo_cookie_sent;   /* Offered a TFO-cookie to the client */
1150         u_int32_t       tcps_tfo_cookie_invalid;/* Received an invalid TFO-cookie */
1151         u_int32_t       tcps_tfo_cookie_req;    /* Cookie requested with the SYN */
1152         u_int32_t       tcps_tfo_cookie_rcv;    /* Cookie received in a SYN/ACK */
1153         u_int32_t       tcps_tfo_syn_data_sent; /* SYN+data+cookie sent */
1154         u_int32_t       tcps_tfo_syn_data_acked;/* SYN+data has been acknowledged */
1155         u_int32_t       tcps_tfo_syn_loss;      /* SYN+TFO has been lost and we fallback */
1156         u_int32_t       tcps_tfo_blackhole;     /* TFO got blackholed by a middlebox. */
1157         u_int32_t       tcps_tfo_cookie_wrong;  /* TFO-cookie we sent was wrong */
1158         u_int32_t       tcps_tfo_no_cookie_rcv; /* We asked for a cookie but didn't get one */
1159         u_int32_t       tcps_tfo_heuristics_disable; /* TFO got disabled due to heuristics */
1160         u_int32_t       tcps_tfo_sndblackhole;  /* TFO got blackholed in the sending direction */
1161         u_int32_t       tcps_mss_to_default;    /* Change MSS to default using link status report */
1162         u_int32_t       tcps_mss_to_medium;     /* Change MSS to medium using link status report */
1163         u_int32_t       tcps_mss_to_low;        /* Change MSS to low using link status report */
1164         u_int32_t       tcps_ecn_fallback_droprst; /* ECN fallback caused by connection drop due to RST */
1165         u_int32_t       tcps_ecn_fallback_droprxmt; /* ECN fallback due to drop after multiple retransmits */
1166 };
1167
1168
1169 struct tcpstat_local {
1170         u_int64_t badformat;
1171         u_int64_t unspecv6;
1172         u_int64_t synfin;
1173         u_int64_t badformatipsec;
1174         u_int64_t noconnnolist;
1175         u_int64_t noconnlist;
1176         u_int64_t listbadsyn;
1177         u_int64_t icmp6unreach;
1178         u_int64_t deprecate6;
1179         u_int64_t ooopacket;
1180         u_int64_t rstinsynrcv;
1181         u_int64_t dospacket;
1182         u_int64_t cleanup;
1183         u_int64_t synwindow;
1184 };
1185
1186 #pragma pack(4)
1187
1188 /*
1189  * TCB structure exported to user-land via sysctl(3).
1190  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
1191  * included.  Not all of our clients do.
1192  */
1193
1194 struct  xtcpcb {
1195         u_int32_t       xt_len;
1196 #ifdef KERNEL_PRIVATE
1197         struct  inpcb_compat    xt_inp;
1198 #else
1199         struct  inpcb   xt_inp;
1200 #endif
1201 #ifdef KERNEL_PRIVATE
1202         struct  otcpcb  xt_tp;
1203 #else
1204         struct  tcpcb   xt_tp;
1205 #endif
1206         struct  xsocket xt_socket;
1207         u_quad_t        xt_alignment_hack;
1208 };
1209
1210
1211 struct  xtcpcb64 {
1212         u_int32_t               xt_len;
1213         struct xinpcb64         xt_inpcb;
1214
1215         u_int64_t t_segq;
1216         int     t_dupacks;              /* consecutive dup acks recd */
1217
1218         int t_timer[TCPT_NTIMERS_EXT];  /* tcp timers */
1219
1220         int     t_state;                /* state of this connection */
1221         u_int   t_flags;
1222
1223         int     t_force;                /* 1 if forcing out a byte */
1224
1225         tcp_seq snd_una;                /* send unacknowledged */
1226         tcp_seq snd_max;                /* highest sequence number sent;
1227                                          * used to recognize retransmits
1228                                          */
1229         tcp_seq snd_nxt;                /* send next */
1230         tcp_seq snd_up;                 /* send urgent pointer */
1231
1232         tcp_seq snd_wl1;                /* window update seg seq number */
1233         tcp_seq snd_wl2;                /* window update seg ack number */
1234         tcp_seq iss;                    /* initial send sequence number */
1235         tcp_seq irs;                    /* initial receive sequence number */
1236
1237         tcp_seq rcv_nxt;                /* receive next */
1238         tcp_seq rcv_adv;                /* advertised window */
1239         u_int32_t rcv_wnd;              /* receive window */
1240         tcp_seq rcv_up;                 /* receive urgent pointer */
1241
1242         u_int32_t snd_wnd;              /* send window */
1243         u_int32_t snd_cwnd;             /* congestion-controlled window */
1244         u_int32_t snd_ssthresh;         /* snd_cwnd size threshold for
1245                                          * for slow start exponential to
1246                                          * linear switch
1247                                          */
1248         u_int   t_maxopd;               /* mss plus options */
1249
1250         u_int32_t t_rcvtime;            /* time at which a packet was received */
1251         u_int32_t t_starttime;          /* time connection was established */
1252         int     t_rtttime;              /* round trip time */
1253         tcp_seq t_rtseq;                /* sequence number being timed */
1254
1255         int     t_rxtcur;               /* current retransmit value (ticks) */
1256         u_int   t_maxseg;               /* maximum segment size */
1257         int     t_srtt;                 /* smoothed round-trip time */
1258         int     t_rttvar;               /* variance in round-trip time */
1259
1260         int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
1261         u_int   t_rttmin;               /* minimum rtt allowed */
1262         u_int32_t t_rttupdated;         /* number of times rtt sampled */
1263         u_int32_t max_sndwnd;           /* largest window peer has offered */
1264
1265         int     t_softerror;            /* possible error not yet reported */
1266 /* out-of-band data */
1267         char    t_oobflags;             /* have some */
1268         char    t_iobc;                 /* input character */
1269 /* RFC 1323 variables */
1270         u_char  snd_scale;              /* window scaling for send window */
1271         u_char  rcv_scale;              /* window scaling for recv window */
1272         u_char  request_r_scale;        /* pending window scaling */
1273         u_char  requested_s_scale;
1274         u_int32_t ts_recent;            /* timestamp echo data */
1275
1276         u_int32_t ts_recent_age;        /* when last updated */
1277         tcp_seq last_ack_sent;
1278 /* RFC 1644 variables */
1279         tcp_cc  cc_send;                /* send connection count */
1280         tcp_cc  cc_recv;                /* receive connection count */
1281         tcp_seq snd_recover;            /* for use in fast recovery */
1282 /* experimental */
1283         u_int32_t snd_cwnd_prev;        /* cwnd prior to retransmit */
1284         u_int32_t snd_ssthresh_prev;    /* ssthresh prior to retransmit */
1285         u_int32_t t_badrxtwin;          /* window for retransmit recovery */
1286
1287         u_quad_t                xt_alignment_hack;
1288 };
1289
1290
1291 #ifdef PRIVATE
1292
1293 struct  xtcpcb_n {
1294         u_int32_t               xt_len;
1295         u_int32_t                       xt_kind;                /* XSO_TCPCB */
1296
1297         u_int64_t t_segq;
1298         int     t_dupacks;              /* consecutive dup acks recd */
1299
1300         int t_timer[TCPT_NTIMERS_EXT];  /* tcp timers */
1301
1302         int     t_state;                /* state of this connection */
1303         u_int   t_flags;
1304
1305         int     t_force;                /* 1 if forcing out a byte */
1306
1307         tcp_seq snd_una;                /* send unacknowledged */
1308         tcp_seq snd_max;                /* highest sequence number sent;
1309                                          * used to recognize retransmits
1310                                          */
1311         tcp_seq snd_nxt;                /* send next */
1312         tcp_seq snd_up;                 /* send urgent pointer */
1313
1314         tcp_seq snd_wl1;                /* window update seg seq number */
1315         tcp_seq snd_wl2;                /* window update seg ack number */
1316         tcp_seq iss;                    /* initial send sequence number */
1317         tcp_seq irs;                    /* initial receive sequence number */
1318
1319         tcp_seq rcv_nxt;                /* receive next */
1320         tcp_seq rcv_adv;                /* advertised window */
1321         u_int32_t rcv_wnd;              /* receive window */
1322         tcp_seq rcv_up;                 /* receive urgent pointer */
1323
1324         u_int32_t snd_wnd;              /* send window */
1325         u_int32_t snd_cwnd;             /* congestion-controlled window */
1326         u_int32_t snd_ssthresh;         /* snd_cwnd size threshold for
1327                                          * for slow start exponential to
1328                                          * linear switch
1329                                          */
1330         u_int   t_maxopd;               /* mss plus options */
1331
1332         u_int32_t t_rcvtime;            /* time at which a packet was received */
1333         u_int32_t t_starttime;          /* time connection was established */
1334         int     t_rtttime;              /* round trip time */
1335         tcp_seq t_rtseq;                /* sequence number being timed */
1336
1337         int     t_rxtcur;               /* current retransmit value (ticks) */
1338         u_int   t_maxseg;               /* maximum segment size */
1339         int     t_srtt;                 /* smoothed round-trip time */
1340         int     t_rttvar;               /* variance in round-trip time */
1341
1342         int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
1343         u_int   t_rttmin;               /* minimum rtt allowed */
1344         u_int32_t t_rttupdated;         /* number of times rtt sampled */
1345         u_int32_t max_sndwnd;           /* largest window peer has offered */
1346
1347         int     t_softerror;            /* possible error not yet reported */
1348         /* out-of-band data */
1349         char    t_oobflags;             /* have some */
1350         char    t_iobc;                 /* input character */
1351         /* RFC 1323 variables */
1352         u_char  snd_scale;              /* window scaling for send window */
1353         u_char  rcv_scale;              /* window scaling for recv window */
1354         u_char  request_r_scale;        /* pending window scaling */
1355         u_char  requested_s_scale;
1356         u_int32_t ts_recent;            /* timestamp echo data */
1357
1358         u_int32_t ts_recent_age;        /* when last updated */
1359         tcp_seq last_ack_sent;
1360         /* RFC 1644 variables */
1361         tcp_cc  cc_send;                /* send connection count */
1362         tcp_cc  cc_recv;                /* receive connection count */
1363         tcp_seq snd_recover;            /* for use in fast recovery */
1364         /* experimental */
1365         u_int32_t snd_cwnd_prev;        /* cwnd prior to retransmit */
1366         u_int32_t snd_ssthresh_prev;    /* ssthresh prior to retransmit */
1367 };
1368
1369 #endif /* PRIVATE */
1370
1371 #pragma pack()
1372
1373 /*
1374  * Names for TCP sysctl objects
1375  */
1376 #define TCPCTL_DO_RFC1323       1       /* use RFC-1323 extensions */
1377 #define TCPCTL_DO_RFC1644       2       /* use RFC-1644 extensions */
1378 #define TCPCTL_MSSDFLT          3       /* MSS default */
1379 #define TCPCTL_STATS            4       /* statistics (read-only) */
1380 #define TCPCTL_RTTDFLT          5       /* default RTT estimate */
1381 #define TCPCTL_KEEPIDLE         6       /* keepalive idle timer */
1382 #define TCPCTL_KEEPINTVL        7       /* interval to send keepalives */
1383 #define TCPCTL_SENDSPACE        8       /* send buffer space */
1384 #define TCPCTL_RECVSPACE        9       /* receive buffer space */
1385 #define TCPCTL_KEEPINIT         10      /* timeout for establishing syn */
1386 #define TCPCTL_PCBLIST          11      /* list of all outstanding PCBs */
1387 #define TCPCTL_DELACKTIME       12      /* time before sending delayed ACK */
1388 #define TCPCTL_V6MSSDFLT        13      /* MSS default for IPv6 */
1389 #define TCPCTL_MAXID            14
1390
1391 #ifdef BSD_KERNEL_PRIVATE
1392 #include <sys/bitstring.h>
1393
1394 #define TCP_PKTLIST_CLEAR(tp) {                                         \
1395         (tp)->t_pktlist_head = (tp)->t_pktlist_tail = NULL;             \
1396         (tp)->t_lastchain = (tp)->t_pktlist_sentlen = 0;                \
1397 }
1398
1399 #define TCPCTL_NAMES { \
1400         { 0, 0 }, \
1401         { "rfc1323", CTLTYPE_INT }, \
1402         { "rfc1644", CTLTYPE_INT }, \
1403         { "mssdflt", CTLTYPE_INT }, \
1404         { "stats", CTLTYPE_STRUCT }, \
1405         { "rttdflt", CTLTYPE_INT }, \
1406         { "keepidle", CTLTYPE_INT }, \
1407         { "keepintvl", CTLTYPE_INT }, \
1408         { "sendspace", CTLTYPE_INT }, \
1409         { "recvspace", CTLTYPE_INT }, \
1410         { "keepinit", CTLTYPE_INT }, \
1411         { "pcblist", CTLTYPE_STRUCT }, \
1412         { "delacktime", CTLTYPE_INT }, \
1413         { "v6mssdflt", CTLTYPE_INT }, \
1414 }
1415
1416 #ifdef SYSCTL_DECL
1417 SYSCTL_DECL(_net_inet_tcp);
1418 #endif /* SYSCTL_DECL */
1419
1420 extern  struct inpcbhead tcb;           /* head of queue of active tcpcb's */
1421 extern  struct inpcbinfo tcbinfo;
1422 extern  struct tcpstat tcpstat; /* tcp statistics */
1423 extern  int tcp_mssdflt;        /* XXX */
1424 extern  int tcp_minmss;
1425 #define TCP_FASTOPEN_SERVER 0x01
1426 #define TCP_FASTOPEN_CLIENT 0x02
1427
1428 extern  int tcp_tfo_halfcnt;
1429 extern  int tcp_tfo_backlog;
1430 extern  int tcp_fastopen;
1431 extern  int tcp_tfo_fallback_min;
1432 extern  int ss_fltsz;
1433 extern  int ss_fltsz_local;
1434 extern  int tcp_do_rfc3390;             /* Calculate ss_fltsz according to RFC 3390 */
1435 extern int tcp_do_rfc1323;
1436 extern int target_qdelay;
1437 extern  u_int32_t tcp_now;              /* for RFC 1323 timestamps */
1438 extern struct timeval tcp_uptime;
1439 extern lck_spin_t *tcp_uptime_lock;
1440 extern  int tcp_delack_enabled;
1441 extern  int tcp_do_sack;        /* SACK enabled/disabled */
1442 extern int tcp_do_rfc3465;
1443 extern int tcp_do_rfc3465_lim2;
1444 extern int maxseg_unacked;
1445 extern int tcp_use_newreno;
1446 extern struct zone *tcp_reass_zone;
1447 extern struct zone *tcp_rxt_seg_zone;
1448 extern int tcp_ecn_outbound;
1449 extern int tcp_ecn_inbound;
1450 extern u_int32_t tcp_autorcvbuf_max;
1451 extern u_int32_t tcp_autorcvbuf_max_ca;
1452 extern u_int32_t tcp_autorcvbuf_inc_shift;
1453
1454 struct protosw;
1455 struct domain;
1456
1457 struct tcp_respond_args {
1458         unsigned int ifscope;
1459         unsigned int nocell:1,
1460                      noexpensive:1,
1461                      awdl_unrestricted:1,
1462                      intcoproc_allowed:1;
1463 };
1464
1465 void     tcp_canceltimers(struct tcpcb *);
1466 struct tcpcb *
1467          tcp_close(struct tcpcb *);
1468 void     tcp_ctlinput(int, struct sockaddr *, void *);
1469 int      tcp_ctloutput(struct socket *, struct sockopt *);
1470 struct tcpcb *
1471          tcp_drop(struct tcpcb *, int);
1472 void     tcp_drain(void);
1473 void     tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt);
1474 struct rmxp_tao *
1475          tcp_gettaocache(struct inpcb *);
1476 void     tcp_init(struct protosw *, struct domain *);
1477 void     tcp_input(struct mbuf *, int);
1478 void     tcp_mss(struct tcpcb *, int, unsigned int);
1479 int      tcp_mssopt(struct tcpcb *);
1480 void     tcp_drop_syn_sent(struct inpcb *, int);
1481 void     tcp_mtudisc(struct inpcb *, int);
1482 struct tcpcb *
1483          tcp_newtcpcb(struct inpcb *);
1484 int      tcp_output(struct tcpcb *);
1485 void     tcp_respond(struct tcpcb *, void *, struct tcphdr *, struct mbuf *,
1486     tcp_seq, tcp_seq, int, struct tcp_respond_args *);
1487 struct rtentry *
1488          tcp_rtlookup(struct inpcb *, unsigned int);
1489 void     tcp_setpersist(struct tcpcb *);
1490 void     tcp_gc(struct inpcbinfo *);
1491 void     tcp_itimer(struct inpcbinfo *ipi);
1492 void     tcp_check_timer_state(struct tcpcb *tp);
1493 void     tcp_run_timerlist(void *arg1, void *arg2);
1494
1495 struct tcptemp *tcp_maketemplate(struct tcpcb *);
1496 void     tcp_fillheaders(struct tcpcb *, void *, void *);
1497 struct tcpcb *tcp_timers(struct tcpcb *, int);
1498 void     tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
1499
1500 void tcp_sack_doack(struct tcpcb *, struct tcpopt *, struct tcphdr *,
1501     u_int32_t *);
1502 extern boolean_t tcp_sack_process_dsack(struct tcpcb *, struct tcpopt *,
1503     struct tcphdr *);
1504 int tcp_detect_bad_rexmt(struct tcpcb *, struct tcphdr *, struct tcpopt *,
1505     u_int32_t rxtime);
1506 void     tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
1507 void     tcp_clean_sackreport(struct tcpcb *tp);
1508 void     tcp_sack_adjust(struct tcpcb *tp);
1509 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1510 void     tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
1511 void     tcp_free_sackholes(struct tcpcb *tp);
1512 int32_t  tcp_sbspace(struct tcpcb *tp);
1513 void     tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp);
1514 void     tcp_set_ecn(struct tcpcb *tp, struct ifnet *ifp);
1515 void     tcp_reset_stretch_ack(struct tcpcb *tp);
1516 extern void tcp_get_ports_used(u_int32_t, int, u_int32_t, bitstr_t *);
1517 uint32_t tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags);
1518 uint32_t tcp_find_anypcb_byaddr(struct ifaddr *ifa);
1519 void     tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so,
1520     u_int32_t maxrcvbuf);
1521 struct bwmeas* tcp_bwmeas_alloc(struct tcpcb *tp);
1522 void tcp_bwmeas_free(struct tcpcb *tp);
1523 extern int32_t timer_diff(uint32_t t1, uint32_t toff1, uint32_t t2, uint32_t toff2);
1524
1525 extern void tcp_set_background_cc(struct socket *);
1526 extern void tcp_set_foreground_cc(struct socket *);
1527 extern void tcp_set_recv_bg(struct socket *);
1528 extern void tcp_clear_recv_bg(struct socket *);
1529 extern boolean_t tcp_sack_byte_islost(struct tcpcb *tp);
1530 #define IS_TCP_RECV_BG(_so)     \
1531         ((_so)->so_flags1 & SOF1_TRAFFIC_MGT_TCP_RECVBG)
1532
1533 #if TRAFFIC_MGT
1534 #define CLEAR_IAJ_STATE(_tp_) (_tp_)->iaj_rcv_ts = 0
1535 void     reset_acc_iaj(struct tcpcb *tp);
1536 #endif /* TRAFFIC_MGT */
1537
1538 int      tcp_lock (struct socket *, int, void *);
1539 int      tcp_unlock (struct socket *, int, void *);
1540 void     calculate_tcp_clock(void);
1541
1542 extern void tcp_keepalive_reset(struct tcpcb *);
1543 extern uint32_t get_base_rtt(struct tcpcb *tp);
1544
1545 #ifdef _KERN_LOCKS_H_
1546 lck_mtx_t *      tcp_getlock (struct socket *, int);
1547 #else
1548 void *   tcp_getlock (struct socket *, int);
1549 #endif
1550
1551 extern  struct pr_usrreqs tcp_usrreqs;
1552 extern  u_int32_t tcp_sendspace;
1553 extern  u_int32_t tcp_recvspace;
1554 tcp_seq tcp_new_isn(struct tcpcb *);
1555
1556 extern int tcp_input_checksum(int, struct mbuf *, struct tcphdr *, int, int);
1557 extern void tcp_getconninfo(struct socket *, struct conninfo_tcp *);
1558 extern void add_to_time_wait(struct tcpcb *, uint32_t delay);
1559 extern void tcp_pmtud_revert_segment_size(struct tcpcb *tp);
1560 extern void tcp_rxtseg_insert(struct tcpcb *, tcp_seq, tcp_seq);
1561 extern struct tcp_rxt_seg *tcp_rxtseg_find(struct tcpcb *, tcp_seq, tcp_seq);
1562 extern void tcp_rxtseg_clean(struct tcpcb *);
1563 extern boolean_t tcp_rxtseg_detect_bad_rexmt(struct tcpcb *, tcp_seq);
1564 extern boolean_t tcp_rxtseg_dsack_for_tlp(struct tcpcb *);
1565 extern u_int32_t tcp_rxtseg_total_size(struct tcpcb *tp);
1566 extern void tcp_rexmt_save_state(struct tcpcb *tp);
1567 extern void tcp_interface_send_probe(u_int16_t if_index_available);
1568 extern void tcp_probe_connectivity(struct ifnet *ifp, u_int32_t enable);
1569 extern void tcp_get_connectivity_status(struct tcpcb *,
1570     struct tcp_conn_status *);
1571
1572 extern void tcp_fill_keepalive_offload_frames(struct ifnet *,
1573     struct ifnet_keepalive_offload_frame *, u_int32_t, size_t, u_int32_t *);
1574
1575 extern boolean_t tfo_enabled(const struct tcpcb *tp);
1576 extern void tcp_disable_tfo(struct tcpcb *tp);
1577 extern void tcp_tfo_gen_cookie(struct inpcb *inp, u_char *out, size_t blk_size);
1578 #define TCP_FASTOPEN_KEYLEN 16
1579 extern errno_t tcp_notify_ack_id_valid(struct tcpcb *, struct socket *, u_int32_t);
1580 extern errno_t tcp_add_notify_ack_marker(struct tcpcb *, u_int32_t);
1581 extern void tcp_notify_ack_free(struct tcpcb *);
1582 extern void tcp_notify_acknowledgement(struct tcpcb *, struct socket *);
1583 extern void tcp_get_notify_ack_count(struct tcpcb *,
1584     struct tcp_notify_ack_complete *);
1585 extern void tcp_get_notify_ack_ids(struct tcpcb *tp,
1586     struct tcp_notify_ack_complete *);
1587 extern void tcp_update_mss_locked(struct socket *, struct ifnet *);
1588
1589 extern int get_tcp_inp_list(struct inpcb **, int, inp_gen_t);
1590 extern bool tcp_notify_ack_active(struct socket *so);
1591
1592 #if MPTCP
1593 extern int mptcp_input_preproc(struct tcpcb *, struct mbuf *, int);
1594 extern void mptcp_output_csum(struct tcpcb *, struct mbuf *, int32_t, unsigned,
1595     u_int64_t, u_int32_t *);
1596 extern int mptcp_adj_mss(struct tcpcb *, boolean_t);
1597 extern void mptcp_insert_rmap(struct tcpcb *, struct mbuf *);
1598 #endif
1599 #endif /* BSD_KERNEL_RPIVATE */
1600
1601 #endif /* _NETINET_TCP_VAR_H_ */