sys/netinet/tcp_var.h

   1 /*
   2  * Copyright (c) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
   3  * Copyright (c) 2003, 2004 The DragonFly Project.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to The DragonFly Project
   6  * by Jeffrey M. Hsu.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of The DragonFly Project nor the names of its
  17  *    contributors may be used to endorse or promote products derived
  18  *    from this software without specific, prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 /*
  35  * Copyright (c) 1982, 1986, 1993, 1994, 1995
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. Neither the name of the University nor the names of its contributors
  47  *    may be used to endorse or promote products derived from this software
  48  *    without specific prior written permission.
  49  *
  50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)tcp_var.h   8.4 (Berkeley) 5/24/95
  63  * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.13 2003/02/03 02:34:07 hsu Exp $
  64  */
  65
  66 #ifndef _NETINET_TCP_VAR_H_
  67 #define _NETINET_TCP_VAR_H_
  68
  69 #ifndef _NETINET_IN_PCB_H_
  70 #include <netinet/in_pcb.h>             /* needed for in_conninfo, inp_gen_t */
  71 #endif
  72 #ifndef _NETINET_TCP_H_
  73 #include <netinet/tcp.h>
  74 #endif
  75
  76 /*
  77  * Kernel variables for tcp.
  78  */
  79 extern int tcp_do_rfc1323;
  80 extern int tcp_low_rtobase;
  81 extern int tcp_ncr_linklocal;
  82 extern int tcp_ncr_rxtthresh_max;
  83 extern int tcp_do_sack;
  84 extern int tcp_do_smartsack;
  85 extern int tcp_do_rescuesack;
  86 extern int tcp_aggressive_rescuesack;
  87 extern int tcp_do_rfc6675;
  88 extern int tcp_rfc6675_rxt;
  89 extern int tcp_aggregate_acks;
  90 extern int tcp_eifel_rtoinc;
  91 extern int tcp_prio_synack;
  92
  93 /* TCP segment queue entry */
  94 struct tseg_qent {
  95         TAILQ_ENTRY(tseg_qent) tqe_q;
  96         int     tqe_len;                /* TCP segment data length */
  97         struct  tcphdr *tqe_th;         /* a pointer to tcp header */
  98         struct  mbuf    *tqe_m;         /* mbuf contains packet */
  99 };
 100 TAILQ_HEAD(tsegqe_head, tseg_qent);
 101 extern int      tcp_reass_maxseg;
 102 extern int      tcp_reass_qsize;
 103 #ifdef MALLOC_DECLARE
 104 MALLOC_DECLARE(M_TSEGQ);
 105 #endif
 106
 107 struct tcptemp {
 108         u_char  tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
 109         struct  tcphdr tt_t;
 110 };
 111
 112 #define tcp6cb          tcpcb  /* for KAME src sync over BSD*'s */
 113
 114 struct raw_sackblock {                          /* covers [start, end) */
 115         tcp_seq rblk_start;
 116         tcp_seq rblk_end;
 117 };
 118
 119 /* maximum number of SACK blocks that will fit in the TCP option space */
 120 #define MAX_SACK_REPORT_BLOCKS  4
 121
 122 TAILQ_HEAD(sackblock_list, sackblock);
 123
 124 struct scoreboard {
 125         int nblocks;
 126         struct sackblock_list sackblocks;
 127         tcp_seq lostseq;                        /* passed SACK lost test */
 128         struct sackblock *lastfound;            /* search hint */
 129         struct sackblock *freecache;            /* one slot free block cache */
 130 };
 131
 132 struct netmsg_tcp_timer;
 133 struct netmsg_base;
 134
 135 struct tcp_pcbport {
 136         struct inpcbport        t_phd;
 137 } __cachealign;
 138
 139 /*
 140  * Tcp control block, one per tcp; fields:
 141  * Organized for 16 byte cacheline efficiency.
 142  */
 143 struct tcpcb {
 144         struct  tsegqe_head t_segq;
 145         int     t_dupacks;              /* consecutive dup acks recd */
 146         int     t_rxtthresh;            /* # dup acks to start fast rxt */
 147         int     tt_cpu;                 /* sanity check the cpu */
 148
 149         struct  tcp_pcbport *t_pcbport; /* per-cpu local port cache for
 150                                          * accept(2)'ed sockets */
 151
 152         struct  tcp_callout *tt_rexmt;  /* retransmit timer */
 153         struct  tcp_callout *tt_persist;/* retransmit persistence */
 154         struct  tcp_callout *tt_keep;   /* keepalive */
 155         struct  tcp_callout *tt_2msl;   /* 2*msl TIME_WAIT timer */
 156         struct  tcp_callout *tt_delack; /* delayed ACK timer */
 157         struct  netmsg_tcp_timer *tt_msg; /* timer message */
 158
 159         struct  netmsg_base *tt_sndmore;/* send more segments (fairsend) */
 160
 161         struct  inpcb *t_inpcb;         /* back pointer to internet pcb */
 162         int     t_state;                /* state of this connection */
 163         u_int   t_flags;
 164 #define TF_ACKNOW       0x00000001      /* ack peer immediately */
 165 #define TF_DELACK       0x00000002      /* ack, but try to delay it */
 166 #define TF_NODELAY      0x00000004      /* don't delay packets to coalesce */
 167 #define TF_NOOPT        0x00000008      /* don't use tcp options */
 168 #define TF_SENTFIN      0x00000010      /* have sent FIN */
 169 #define TF_REQ_SCALE    0x00000020      /* have/will request window scaling */
 170 #define TF_RCVD_SCALE   0x00000040      /* other side has requested scaling */
 171 #define TF_REQ_TSTMP    0x00000080      /* have/will request timestamps */
 172 #define TF_RCVD_TSTMP   0x00000100      /* a timestamp was received in SYN */
 173 #define TF_SACK_PERMITTED 0x00000200    /* other side said I could SACK */
 174 #define TF_NEEDSYN      0x00000400      /* send SYN (implicit state) */
 175 #define TF_NEEDFIN      0x00000800      /* send FIN (implicit state) */
 176 #define TF_NOPUSH       0x00001000      /* don't push */
 177 #define TF_LISTEN       0x00002000      /* listen(2) has been called */
 178 #define TF_SIGNATURE    0x00004000      /* require MD5 digests (RFC2385) */
 179 #define TF_NCR          0x00008000      /* Non-Congestion Robustness RFC4653 */
 180 #define TF_MORETOCOME   0x00010000      /* More data to be appended to sock */
 181 #define TF_SAWFIN       0x00020000      /* FIN has been seen */
 182 #define TF_LASTIDLE     0x00040000      /* connection was previously idle */
 183 #define TF_RXWIN0SENT   0x00080000      /* sent a receiver win 0 in response */
 184 #define TF_FASTRECOVERY 0x00100000      /* in Fast Recovery */
 185 #define TF_QUEDFIN      0x00200000      /* FIN has been received */
 186 #define TF_XMITNOW      0x00400000      /* Temporarily override Nagle */
 187 #define TF_UNUSED008    0x00800000
 188 #define TF_UNUSED009    0x01000000
 189 #define TF_FORCE        0x02000000      /* Set if forcing out a byte */
 190 #define TF_ONOUTPUTQ    0x04000000      /* on t_outputq list */
 191 #define TF_FAIRSEND     0x08000000      /* fairsend is requested */
 192 #define TF_UNUSED003    0x10000000
 193 #define TF_UNUSED004    0x20000000
 194 #define TF_KEEPALIVE    0x40000000      /* temporary keepalive */
 195 #define TF_RXRESIZED    0x80000000      /* rcvbuf was resized */
 196         tcp_seq snd_up;                 /* send urgent pointer */
 197         u_long  snd_last;               /* time last data were sent */
 198
 199         tcp_seq snd_una;                /* send unacknowledged */
 200         tcp_seq snd_recover;            /* for use with Fast Recovery */
 201         tcp_seq snd_max;                /* highest sequence number sent;
 202                                          * used to recognize retransmits */
 203         tcp_seq snd_nxt;                /* send next */
 204
 205         tcp_seq snd_wl1;                /* window update seg seq number */
 206         tcp_seq snd_wl2;                /* window update seg ack number */
 207         tcp_seq iss;                    /* initial send sequence number */
 208         tcp_seq irs;                    /* initial receive sequence number */
 209
 210         tcp_seq rcv_nxt;                /* receive next */
 211         tcp_seq rcv_adv;                /* advertised window */
 212         u_long  rcv_wnd;                /* receive window */
 213         tcp_seq rcv_up;                 /* receive urgent pointer */
 214
 215         u_long  snd_wnd;                /* send window */
 216         u_long  snd_cwnd;               /* congestion-controlled window */
 217         u_long  snd_wacked;             /* bytes acked in one send window */
 218         u_long  snd_ssthresh;           /* snd_cwnd size threshold for
 219                                          * for slow start exponential to
 220                                          * linear switch */
 221
 222         int     t_rxtcur;               /* current retransmit value (ticks) */
 223         u_int   t_maxseg;               /* maximum segment size */
 224         int     t_srtt;                 /* smoothed round-trip time */
 225         int     t_rttvar;               /* variance in round-trip time */
 226
 227         u_int   t_maxopd;               /* mss plus options */
 228
 229         u_long  t_rcvtime;              /* reception inactivity time */
 230         u_long  t_starttime;            /* time connection was established */
 231         int     t_rtttime;              /* round trip time */
 232         tcp_seq t_rtseq;                /* sequence number being timed */
 233
 234         int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
 235         u_int   t_rttmin;               /* minimum rtt allowed */
 236         u_int   t_rttbest;              /* best rtt we've seen */
 237         u_long  t_rttupdated;           /* number of times rtt sampled */
 238         u_long  max_sndwnd;             /* largest window peer has offered */
 239
 240         int     t_softerror;            /* possible error not yet reported */
 241 /* out-of-band data */
 242         char    t_oobflags;             /* have some */
 243         char    t_iobc;                 /* input character */
 244 #define TCPOOB_HAVEDATA 0x01
 245 #define TCPOOB_HADDATA  0x02
 246
 247 /* RFC 1323 variables */
 248         u_char  snd_scale;              /* window scaling for send window */
 249         u_char  rcv_scale;              /* window scaling for recv window */
 250         u_char  request_r_scale;        /* pending window scaling */
 251         u_long  ts_recent;              /* timestamp echo data */
 252
 253         u_long  ts_recent_age;          /* when last updated */
 254         tcp_seq last_ack_sent;
 255
 256 /* experimental */
 257         u_int   rxt_flags;
 258 #define TRXT_F_REBASERTO        0x0001  /* Recalculate RTO based on new RTT */
 259 #define TRXT_F_WASFRECOVERY     0x0002  /* was in Fast Recovery */
 260 #define TRXT_F_FIRSTACCACK      0x0004  /* Look for 1st acceptable ACK. */
 261 #define TRXT_F_FASTREXMT        0x0008  /* Did Fast Retransmit. */
 262 #define TRXT_F_EARLYREXMT       0x0010  /* Did Early (Fast) Retransmit. */
 263         int     t_srtt_prev;            /* adjusted SRTT prior to retransmit */
 264         int     t_rttvar_prev;          /* RTTVAR prior to retransmit */
 265         int     t_rxtcur_prev;          /* rexmt timeout prior to retransmit */
 266         tcp_seq snd_max_prev;           /* SND_MAX prior to retransmit */
 267         u_long  snd_cwnd_prev;          /* cwnd prior to retransmit */
 268         u_long  snd_wacked_prev;        /* prior bytes acked in send window */
 269         u_long  snd_ssthresh_prev;      /* ssthresh prior to retransmit */
 270         tcp_seq snd_recover_prev;       /* snd_recover prior to retransmit */
 271         u_long  t_badrxtwin;            /* window for retransmit recovery */
 272         u_long  t_rexmtTS;              /* timestamp of last retransmit */
 273         u_char  snd_limited;            /* segments limited transmitted */
 274
 275         u_int   sack_flags;
 276 #define TSACK_F_SACKRESCUED     0x0001  /* sent rescue SACK recovery data */
 277 #define TSACK_F_DUPSEG          0x0002  /* last seg a duplicate */
 278 #define TSACK_F_ENCLOSESEG      0x0004  /* enclosing SACK block */
 279 #define TSACK_F_SACKLEFT        0x0008  /* send SACK blocks from left side */
 280         tcp_seq rexmt_high;             /* highest seq # retransmitted + 1 */
 281         tcp_seq rexmt_rescue;           /* rescue SACKED sequence number */
 282         tcp_seq snd_max_rexmt;          /* snd_max when rexmting snd_una */
 283         struct scoreboard scb;          /* sack scoreboard */
 284         struct raw_sackblock reportblk; /* incoming segment or D-SACK block */
 285         struct raw_sackblock encloseblk;
 286         int     nsackhistory;
 287         struct raw_sackblock sackhistory[MAX_SACK_REPORT_BLOCKS]; /* reported */
 288
 289         TAILQ_ENTRY(tcpcb) t_outputq;   /* tcp_output needed list */
 290
 291         /* bandwith limitation */
 292         u_long  snd_bandwidth;          /* calculated bandwidth or 0 */
 293         u_long  snd_bwnd;               /* bandwidth-controlled window */
 294         int     t_bw_rtttime;           /* used for bandwidth calculation */
 295         tcp_seq t_bw_rtseq;             /* used for bandwidth calculation */
 296
 297 /* anti DoS counters */
 298         u_long  rcv_second;             /* start of interval second */
 299         u_long  rcv_pps;                /* received packets per second */
 300         u_long  rcv_byps;               /* received bytes per second */
 301
 302         u_int32_t       rfbuf_ts;       /* recv buffer autoscaling timestamp */
 303         int     rfbuf_cnt;              /* recv buffer autoscaling byte count */
 304
 305         int     t_keepinit;             /* time to establish connection */
 306
 307         int     t_keepidle;             /* time before keepalive probes begin */
 308         int     t_keepintvl;            /* time between keepalive probes */
 309         int     t_keepcnt;              /* maximum number of keepalive probes */
 310         int     t_maxidle;              /* time to drop after starting probes */
 311
 312         int     t_rxtsyn;               /* time spent in SYN or SYN|ACK rexmt */
 313 };
 314
 315 #define IN_FASTRECOVERY(tp)     (tp->t_flags & TF_FASTRECOVERY)
 316 #define ENTER_FASTRECOVERY(tp)  tp->t_flags |= TF_FASTRECOVERY
 317 #define EXIT_FASTRECOVERY(tp)   tp->t_flags &= ~TF_FASTRECOVERY
 318
 319 #ifdef TCP_SIGNATURE
 320 /*
 321  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
 322  * for SADB verification and lookup.
 323  */
 324 #define TCP_SIGLEN      16      /* length of computed digest in bytes */
 325 #define TCP_KEYLEN_MIN  1       /* minimum length of TCP-MD5 key */
 326 #define TCP_KEYLEN_MAX  80      /* maximum length of TCP-MD5 key */
 327 /*
 328  * Only a single SA per host may be specified at this time. An SPI is
 329  * needed in order for the KEY_ALLOCSA() lookup to work.
 330  */
 331 #define TCP_SIG_SPI     0x1000
 332 #endif /* TCP_SIGNATURE */
 333
 334 /*
 335  * TCP statistics.
 336  *
 337  * NOTE: Make sure this struct's size is multiple cache line size.
 338  */
 339 struct tcp_stats {
 340         u_long  tcps_connattempt;       /* connections initiated */
 341         u_long  tcps_accepts;           /* connections accepted */
 342         u_long  tcps_connects;          /* connections established */
 343         u_long  tcps_drops;             /* connections dropped */
 344         u_long  tcps_conndrops;         /* embryonic connections dropped */
 345         u_long  tcps_minmssdrops;       /* average minmss too low drops */
 346         u_long  tcps_closed;            /* conn. closed (includes drops) */
 347         u_long  tcps_segstimed;         /* segs where we tried to get rtt */
 348         u_long  tcps_rttupdated;        /* times we succeeded */
 349         u_long  tcps_delack;            /* delayed acks sent */
 350         u_long  tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
 351         u_long  tcps_rexmttimeo;        /* retransmit timeouts */
 352         u_long  tcps_persisttimeo;      /* persist timeouts */
 353         u_long  tcps_keeptimeo;         /* keepalive timeouts */
 354         u_long  tcps_keepprobe;         /* keepalive probes sent */
 355         u_long  tcps_keepdrops;         /* connections dropped in keepalive */
 356
 357         u_long  tcps_sndtotal;          /* total packets sent */
 358         u_long  tcps_sndpack;           /* data packets sent */
 359         u_long  tcps_sndbyte;           /* data bytes sent */
 360         u_long  tcps_sndrexmitpack;     /* data packets retransmitted */
 361         u_long  tcps_sndrexmitbyte;     /* data bytes retransmitted */
 362         u_long  tcps_sndsackrtopack;    /* packets sent by SACK after RTO */
 363         u_long  tcps_sndsackrtobyte;    /* bytes sent by SACK after RTO */
 364         u_long  tcps_sndfastrexmit;     /* Fast Retransmissions */
 365         u_long  tcps_sndearlyrexmit;    /* early Fast Retransmissions */
 366         u_long  tcps_sndlimited;        /* Limited Transmit packets */
 367         u_long  tcps_sndrtobad;         /* spurious RTO retransmissions */
 368         u_long  tcps_sndfastrexmitbad;  /* spurious Fast Retransmissions */
 369         u_long  tcps_sndearlyrexmitbad; /* spurious early Fast Retransmissions,
 370                                            a subset of tcps_sndfastrexmitbad */
 371         u_long  tcps_eifeldetected;     /* Eifel-detected spurious rexmits */
 372         u_long  tcps_rttcantdetect;     /* Eifel but not 1/2 RTT-detectable */
 373         u_long  tcps_rttdetected;       /* RTT-detected spurious RTO rexmits */
 374         u_long  tcps_sndacks;           /* ack-only packets sent */
 375         u_long  tcps_sndprobe;          /* window probes sent */
 376         u_long  tcps_sndurg;            /* packets sent with URG only */
 377         u_long  tcps_sndwinup;          /* window update-only packets sent */
 378         u_long  tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
 379         u_long  tcps_sndsackpack;       /* packets sent by SACK recovery alg */
 380         u_long  tcps_sndsackbyte;       /* bytes sent by SACK recovery */
 381         u_long  tcps_snduna3;           /* re-retransmit snd_una on 3 new seg */
 382         u_long  tcps_snduna1;           /* re-retransmit snd_una on 1 new seg */
 383         u_long  tcps_sndsackopt;        /* SACK options sent */
 384         u_long  tcps_snddsackopt;       /* D-SACK options sent */
 385         u_long  tcps_sndidle;           /* sending idle detected */
 386         u_long  tcps_sackrescue;        /* SACK rescue data packets sent */
 387         u_long  tcps_sackrescue_try;    /* SACK rescues attempted */
 388         u_long  tcps_eifelresponse;     /* Eifel responses */
 389
 390         u_long  tcps_rcvtotal;          /* total packets received */
 391         u_long  tcps_rcvpack;           /* packets received in sequence */
 392         u_long  tcps_rcvbyte;           /* bytes received in sequence */
 393         u_long  tcps_rcvbadsum;         /* packets received with ccksum errs */
 394         u_long  tcps_rcvbadoff;         /* packets received with bad offset */
 395         u_long  tcps_rcvmemdrop;        /* packets dropped for lack of memory */
 396         u_long  tcps_rcvshort;          /* packets received too short */
 397         u_long  tcps_rcvduppack;        /* duplicate-only packets received */
 398         u_long  tcps_rcvdupbyte;        /* duplicate-only bytes received */
 399         u_long  tcps_rcvpartduppack;    /* packets with some duplicate data */
 400         u_long  tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
 401         u_long  tcps_rcvoopack;         /* out-of-order packets received */
 402         u_long  tcps_rcvoobyte;         /* out-of-order bytes received */
 403         u_long  tcps_rcvpackafterwin;   /* packets with data after window */
 404         u_long  tcps_rcvbyteafterwin;   /* bytes rcvd after window */
 405         u_long  tcps_rcvafterclose;     /* packets rcvd after "close" */
 406         u_long  tcps_rcvwinprobe;       /* rcvd window probe packets */
 407         u_long  tcps_rcvdupack;         /* rcvd duplicate acks */
 408         u_long  tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
 409         u_long  tcps_rcvackpack;        /* rcvd ack packets */
 410         u_long  tcps_rcvackbyte;        /* bytes acked by rcvd acks */
 411         u_long  tcps_rcvwinupd;         /* rcvd window update packets */
 412         u_long  tcps_pawsdrop;          /* segments dropped due to PAWS */
 413         u_long  tcps_pawsaccept;        /* segments accepted, PAWS tolerance */
 414         u_long  tcps_predack;           /* times hdr predict ok for acks */
 415         u_long  tcps_preddat;           /* times hdr predict ok for data pkts */
 416         u_long  tcps_pcbcachemiss;
 417         u_long  tcps_cachedrtt;         /* times cached RTT in route updated */
 418         u_long  tcps_cachedrttvar;      /* times cached rttvar updated */
 419         u_long  tcps_cachedssthresh;    /* times cached ssthresh updated */
 420         u_long  tcps_usedrtt;           /* times RTT initialized from route */
 421         u_long  tcps_usedrttvar;        /* times RTTVAR initialized from rt */
 422         u_long  tcps_usedssthresh;      /* times ssthresh initialized from rt*/
 423         u_long  tcps_persistdrop;       /* timeout in persist state */
 424         u_long  tcps_badsyn;            /* bogus SYN, e.g. premature ACK */
 425         u_long  tcps_mturesent;         /* resends due to MTU discovery */
 426         u_long  tcps_listendrop;        /* listen queue overflows */
 427         u_long  tcps_rcvbadsackopt;     /* rcvd illegal SACK options */
 428         u_long  tcps_sackrenege;        /* times other side reneged */
 429
 430         u_long  tcps_sacksbupdate;      /* times SACK scoreboard updated */
 431         u_long  tcps_sacksboverflow;    /* times SACK scoreboard overflowed */
 432         u_long  tcps_sacksbreused;      /* times SACK sb-block reused */
 433         u_long  tcps_sacksbfailed;      /* times SACK sb update failed */
 434         u_long  tcps_sacksbfast;        /* times SACK sb-block uses cache */
 435
 436         u_long  tcps_sc_added;          /* entry added to syncache */
 437         u_long  tcps_sc_retransmitted;  /* syncache entry was retransmitted */
 438         u_long  tcps_sc_dupsyn;         /* duplicate SYN packet */
 439         u_long  tcps_sc_dropped;        /* could not reply to packet */
 440         u_long  tcps_sc_completed;      /* successful extraction of entry */
 441         u_long  tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
 442         u_long  tcps_sc_cacheoverflow;  /* syncache cache limit hit */
 443         u_long  tcps_sc_reset;          /* RST removed entry from syncache */
 444         u_long  tcps_sc_stale;          /* timed out or listen socket gone */
 445         u_long  tcps_sc_aborted;        /* syncache entry aborted */
 446         u_long  tcps_sc_badack;         /* removed due to bad ACK */
 447         u_long  tcps_sc_unreach;        /* ICMP unreachable received */
 448         u_long  tcps_sc_zonefail;       /* zalloc() failed */
 449         u_long  tcps_sc_sendcookie;     /* SYN cookie sent */
 450         u_long  tcps_sc_recvcookie;     /* SYN cookie received */
 451
 452         u_long  tcps_pad[6];            /* pad to cache line size (64B) */
 453 };
 454 #ifdef _KERNEL
 455 CTASSERT((sizeof(struct tcp_stats) & __VM_CACHELINE_MASK) == 0);
 456 #endif
 457
 458 #ifdef _KERNEL
 459
 460 #ifndef _NETINET_TCP_FSM_H_
 461 #include <netinet/tcp_fsm.h>
 462 #endif
 463
 464 struct tcp_state_count {
 465         u_long  tcps_count[TCP_NSTATES];
 466 } __cachealign;
 467
 468 #define tcpstat tcpstats_percpu[mycpuid]
 469
 470 struct sockopt;
 471
 472 extern struct tcp_stats         tcpstats_percpu[MAXCPU];
 473 extern struct tcp_state_count   tcpstate_count[MAXCPU];
 474
 475 static const int tcprexmtthresh = 3;
 476 #endif
 477
 478 /*
 479  * Structure to hold TCP options that are only used during segment
 480  * processing (in tcp_input), but not held in the tcpcb.
 481  * It's basically used to reduce the number of parameters
 482  * to tcp_dooptions.
 483  */
 484 struct tcpopt {
 485         u_int           to_flags;       /* which options are present */
 486 #define TOF_TS                  0x0001  /* timestamp */
 487 #define TOF_MSS                 0x0010
 488 #define TOF_SCALE               0x0020
 489 #define TOF_SACK_PERMITTED      0x0040
 490 #define TOF_SACK                0x0080
 491 #define TOF_SIGNATURE           0x0100  /* signature option present */
 492 #define TOF_SIGLEN              0x0200  /* signature length valid (RFC2385) */
 493 #define TOF_DSACK               0x0400  /* D-SACK */
 494 #define TOF_SACK_REDUNDANT      0x0800  /* all SACK blocks are known */
 495         u_int32_t       to_tsval;
 496         u_int32_t       to_tsecr;
 497         u_int16_t       to_mss;
 498         u_int8_t        to_requested_s_scale;
 499         u_int8_t        to_nsackblocks;
 500         struct raw_sackblock *to_sackblocks;
 501 };
 502
 503 struct syncache {
 504         inp_gen_t       sc_inp_gencnt;          /* pointer check */
 505         struct          tcpcb *sc_tp;           /* tcb for listening socket */
 506         struct          mbuf *sc_ipopts;        /* source route */
 507         struct          in_conninfo sc_inc;     /* addresses */
 508 #define sc_route        sc_inc.inc_route
 509 #define sc_route6       sc_inc.inc6_route
 510         u_int32_t       sc_tsrecent;
 511         uint16_t        sc_hashval;             /* connection hash */
 512         uint16_t        sc_pad;                 /* explicit padding */
 513         tcp_seq         sc_irs;                 /* seq from peer */
 514         tcp_seq         sc_iss;                 /* our ISS */
 515         u_long          sc_rxttime;             /* retransmit time */
 516         u_int16_t       sc_rxtslot;             /* retransmit counter */
 517         u_int16_t       sc_peer_mss;            /* peer's MSS */
 518         u_int16_t       sc_wnd;                 /* advertised window */
 519         u_int8_t        sc_requested_s_scale:4,
 520                         sc_request_r_scale:4;
 521         u_int8_t        sc_flags;
 522 #define SCF_NOOPT               0x01            /* no TCP options */
 523 #define SCF_WINSCALE            0x02            /* negotiated window scaling */
 524 #define SCF_TIMESTAMP           0x04            /* negotiated timestamps */
 525 #define SCF_HASH                0x08            /* sc_hashval is valid */
 526 #define SCF_UNREACH             0x10            /* icmp unreachable received */
 527 #define SCF_SACK_PERMITTED      0x20            /* saw SACK permitted option */
 528 #define SCF_SIGNATURE           0x40            /* send MD5 digests */
 529         int             sc_rxtused;             /* time spent in SYN|ACK rxt */
 530         u_long          sc_sndwnd;              /* send window */
 531         TAILQ_ENTRY(syncache) sc_hash;
 532         TAILQ_ENTRY(syncache) sc_timerq;
 533 };
 534
 535 struct syncache_head {
 536         TAILQ_HEAD(, syncache)  sch_bucket;
 537         u_int           sch_length;
 538 };
 539
 540 #define intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
 541 #define sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
 542
 543 /*
 544  * The smoothed round-trip time and estimated variance
 545  * are stored as fixed point numbers scaled by the values below.
 546  * For convenience, these scales are also used in smoothing the average
 547  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
 548  * With these scales, srtt has 3 bits to the right of the binary point,
 549  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
 550  * binary point, and is smoothed with an ALPHA of 0.75.
 551  */
 552 #define TCP_RTT_SCALE           32      /* multiplier for srtt; 3 bits frac. */
 553 #define TCP_RTT_SHIFT           5       /* shift for srtt; 3 bits frac. */
 554 #define TCP_RTTVAR_SCALE        16      /* multiplier for rttvar; 2 bits */
 555 #define TCP_RTTVAR_SHIFT        4       /* shift for rttvar; 2 bits */
 556 #define TCP_DELTA_SHIFT         2       /* see tcp_input.c */
 557
 558 /*
 559  * The initial retransmission should happen at rtt + 4 * rttvar.
 560  * Because of the way we do the smoothing, srtt and rttvar
 561  * will each average +1/2 tick of bias.  When we compute
 562  * the retransmit timer, we want 1/2 tick of rounding and
 563  * 1 extra tick because of +-1/2 tick uncertainty in the
 564  * firing of the timer.  The bias will give us exactly the
 565  * 1.5 tick we need.  But, because the bias is
 566  * statistical, we have to test that we don't drop below
 567  * the minimum feasible timer (which is 2 ticks).
 568  * This version of the macro adapted from a paper by Lawrence
 569  * Brakmo and Larry Peterson which outlines a problem caused
 570  * by insufficient precision in the original implementation,
 571  * which results in inappropriately large RTO values for very
 572  * fast networks.
 573  */
 574 #define TCP_REXMTVAL(tp) \
 575         max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
 576           + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
 577
 578 /*
 579  * TCB structure exported to user-land via sysctl(3).
 580  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
 581  * included.  Not all of our clients do.
 582  */
 583 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
 584 struct  xtcpcb {
 585         size_t  xt_len;
 586         struct  inpcb   xt_inp;
 587         struct  tcpcb   xt_tp;
 588         struct  xsocket xt_socket;
 589         u_quad_t        xt_alignment_hack;
 590 };
 591 #endif
 592
 593 /*
 594  * Names for TCP sysctl objects
 595  */
 596 #define TCPCTL_DO_RFC1323       1       /* use RFC-1323 extensions */
 597 /* 2 was TCPCTL_DO_RFC1644 */
 598 #define TCPCTL_MSSDFLT          3       /* MSS default */
 599 #define TCPCTL_STATS            4       /* statistics (read-only) */
 600 #define TCPCTL_RTTDFLT          5       /* default RTT estimate */
 601 #define TCPCTL_KEEPIDLE         6       /* keepalive idle timer */
 602 #define TCPCTL_KEEPINTVL        7       /* interval to send keepalives */
 603 #define TCPCTL_SENDSPACE        8       /* send buffer space */
 604 #define TCPCTL_RECVSPACE        9       /* receive buffer space */
 605 #define TCPCTL_KEEPINIT         10      /* timeout for establishing syn */
 606 #define TCPCTL_PCBLIST          11      /* list of all outstanding PCBs */
 607 #define TCPCTL_DELACKTIME       12      /* time before sending delayed ACK */
 608 #define TCPCTL_V6MSSDFLT        13      /* MSS default for IPv6 */
 609 #define TCPCTL_MAXID            14
 610
 611 #define TCPCTL_NAMES { \
 612         { 0, 0 }, \
 613         { "rfc1323", CTLTYPE_INT }, \
 614           { "reserved", CTLTYPE_INT},   /* was rfc1644 */       \
 615         { "mssdflt", CTLTYPE_INT }, \
 616         { "stats", CTLTYPE_STRUCT }, \
 617         { "rttdflt", CTLTYPE_INT }, \
 618         { "keepidle", CTLTYPE_INT }, \
 619         { "keepintvl", CTLTYPE_INT }, \
 620         { "sendspace", CTLTYPE_INT }, \
 621         { "recvspace", CTLTYPE_INT }, \
 622         { "keepinit", CTLTYPE_INT }, \
 623         { "pcblist", CTLTYPE_STRUCT }, \
 624         { "delacktime", CTLTYPE_INT }, \
 625         { "v6mssdflt", CTLTYPE_INT }, \
 626 }
 627
 628 #ifdef _KERNEL
 629 #ifdef SYSCTL_DECL
 630 SYSCTL_DECL(_net_inet_tcp);
 631 #endif
 632
 633 #define TCP_DO_SACK(tp)         ((tp)->t_flags & TF_SACK_PERMITTED)
 634 #define TCP_DO_NCR(tp)          (((tp)->t_flags & TF_NCR) && TCP_DO_SACK((tp)))
 635 #define TCP_SACK_BLKEND(len, thflags) \
 636         ((len) + (((thflags) & TH_FIN) != 0))
 637
 638 struct tcpcbackq {
 639         TAILQ_HEAD(, tcpcb)     head;
 640 } __cachealign;
 641
 642 extern  struct inpcbinfo tcbinfo[];
 643 extern  struct tcpcbackq tcpcbackq[];
 644
 645 extern  int tcp_mssdflt;        /* XXX */
 646 extern  int tcp_minmss;
 647 extern  int tcp_delack_enabled;
 648 extern  int path_mtu_discovery;
 649
 650 struct ip;
 651 union netmsg;
 652
 653 int      tcp_addrcpu(in_addr_t faddr, in_port_t fport,
 654             in_addr_t laddr, in_port_t lport);
 655 int      tcp_addrhash(in_addr_t faddr, in_port_t fport,
 656             in_addr_t laddr, in_port_t lport);
 657 struct lwkt_port *
 658         tcp_addrport(in_addr_t faddr, in_port_t fport,
 659             in_addr_t laddr, in_port_t lport);
 660 struct lwkt_port *tcp_addrport0(void);
 661 void     tcp_canceltimers (struct tcpcb *);
 662 struct tcpcb *
 663          tcp_close (struct tcpcb *);
 664 void     tcp_ctlinput(union netmsg *);
 665 void     tcp_ctloutput(union netmsg *);
 666 struct netmsg_pr_ctloutput *tcp_ctloutmsg(struct sockopt *);
 667 inp_notify_t tcp_get_inpnotify(int cmd, const struct sockaddr *sa,
 668             int *arg, struct ip **ip0, int *cpuid);
 669 struct tcpcb *
 670          tcp_drop (struct tcpcb *, int);
 671 void     tcp_drain (void);
 672 void     tcp_init (void);
 673 void     tcp_thread_init (void);
 674 int      tcp_input (struct mbuf **, int *, int);
 675 void     tcp_rmx_init (struct tcpcb *, int);
 676 int      tcp_mssopt (struct tcpcb *);
 677 void     tcp_drop_syn_sent (struct inpcb *, int);
 678 void     tcp_mtudisc (struct inpcb *, int);
 679 void     tcp_newtcpcb (struct inpcb *);
 680 int      tcp_output(struct tcpcb *);
 681 int      tcp_output_fair(struct tcpcb *);
 682 void     tcp_output_init(struct tcpcb *);
 683 void     tcp_output_cancel(struct tcpcb *);
 684 boolean_t
 685          tcp_output_pending(struct tcpcb *);
 686 void     tcp_quench (struct inpcb *, int);
 687 void     tcp_respond (struct tcpcb *, void *,
 688             struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
 689 struct rtentry *
 690          tcp_rtlookup (struct in_conninfo *);
 691 int      tcp_sack_bytes_below(const struct scoreboard *scb, tcp_seq seq);
 692 void     tcp_sack_destroy(struct scoreboard *scb);
 693 void     tcp_sack_discard(struct tcpcb *tp);
 694 void     tcp_sack_report_cleanup(struct tcpcb *tp);
 695 boolean_t
 696          tcp_sack_report_needed(const struct tcpcb *tp);
 697 int      tcp_sack_ndsack_blocks(const struct raw_sackblock *blocks,
 698             const int numblocks, tcp_seq snd_una);
 699 void     tcp_sack_fill_report(struct tcpcb *tp, u_char *opt, u_int *plen);
 700 boolean_t
 701          tcp_sack_has_sacked(const struct scoreboard *scb, u_int amount);
 702 void     tcp_sack_tcpcb_init(struct tcpcb *tp);
 703 uint32_t tcp_sack_compute_pipe(const struct tcpcb *tp);
 704 boolean_t
 705          tcp_sack_nextseg(struct tcpcb *tp, tcp_seq *nextrexmt, uint32_t *len,
 706                           boolean_t *rescue);
 707 boolean_t
 708          tcp_sack_islost(const struct scoreboard *scb, tcp_seq seq);
 709 void     tcp_sack_update_lostseq(struct scoreboard *scb, tcp_seq snd_una,
 710             u_int maxseg, int rxtthresh);
 711 #ifdef later
 712 void     tcp_sack_revert_scoreboard(struct scoreboard *scb, tcp_seq snd_una,
 713                                     u_int maxseg);
 714 void     tcp_sack_save_scoreboard(struct scoreboard *scb);
 715 #endif
 716 void     tcp_sack_skip_sacked(struct scoreboard *scb, tcp_seq *prexmt);
 717 uint32_t tcp_sack_first_unsacked_len(const struct tcpcb *tp);
 718 void     tcp_sack_update_scoreboard(struct tcpcb *tp, struct tcpopt *to);
 719 void     tcp_save_congestion_state(struct tcpcb *tp);
 720 void     tcp_revert_congestion_state(struct tcpcb *tp);
 721 void     tcp_setpersist (struct tcpcb *);
 722 struct tcptemp *tcp_maketemplate (struct tcpcb *);
 723 void     tcp_freetemplate (struct tcptemp *);
 724 void     tcp_fillheaders (struct tcpcb *, void *, void *, boolean_t);
 725 struct lwkt_port *
 726          tcp_soport(struct socket *, struct sockaddr *, struct mbuf **);
 727 struct lwkt_port *
 728          tcp_ctlport(int, struct sockaddr *, void *, int *);
 729 struct lwkt_port *
 730          tcp_initport(void);
 731 struct tcpcb *
 732          tcp_timers (struct tcpcb *, int);
 733 void     tcp_trace (short, short, struct tcpcb *, void *, struct tcphdr *,
 734                         int);
 735 void     tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
 736 u_long   tcp_initial_window(struct tcpcb *tp);
 737 void     tcp_timer_keep_activity(struct tcpcb *tp, int thflags);
 738 void     syncache_init(void);
 739 void     syncache_unreach(struct in_conninfo *, const struct tcphdr *);
 740 int      syncache_expand(struct in_conninfo *, struct tcphdr *,
 741              struct socket **, struct mbuf *);
 742 int      syncache_add(struct in_conninfo *, struct tcpopt *,
 743              struct tcphdr *, struct socket *, struct mbuf *);
 744 void     syncache_chkrst(struct in_conninfo *, struct tcphdr *);
 745 void     syncache_badack(struct in_conninfo *);
 746 void     syncache_destroy(struct tcpcb *tp, struct tcpcb *new_tp);
 747
 748 #ifdef TCP_SIGNATURE
 749 int tcpsignature_apply(void *fstate, void *data, unsigned int len);
 750 int tcpsignature_compute(struct mbuf *m, int len, int tcpoptlen,
 751                 u_char *buf, u_int direction);
 752 #endif /* TCP_SIGNATURE */
 753
 754 extern  struct pr_usrreqs tcp_usrreqs;
 755 extern  u_long tcp_sendspace;
 756 extern  u_long tcp_recvspace;
 757 tcp_seq tcp_new_isn (struct tcpcb *);
 758
 759 void    tcp_pcbport_create(struct tcpcb *);
 760 void    tcp_pcbport_destroy(struct tcpcb *);
 761 void    tcp_pcbport_merge_oncpu(struct tcpcb *);
 762
 763 static __inline void
 764 tcp_pcbport_insert(struct tcpcb *ltp, struct inpcb *inp)
 765 {
 766         struct inpcbport *phd;
 767         int cpu;
 768
 769         if (inp->inp_lport != ltp->t_inpcb->inp_lport) {
 770                 /*
 771                  * This could happen with 'ipfw forward'.
 772                  */
 773                 in_pcbinsporthash_lport(inp);
 774                 return;
 775         }
 776
 777         cpu = mycpuid;
 778         KASSERT(cpu < netisr_ncpus, ("invalid cpu%d", cpu));
 779         phd = &ltp->t_pcbport[cpu].t_phd;
 780
 781         /*
 782          * NOTE:
 783          * Set inp_porthash NULL and set inp_phd properly,
 784          * so that tcp_pcbport_remove() could tell that this
 785          * inpcb is on the listen tcpcb per-cpu port cache.
 786          */
 787         inp->inp_porthash = NULL;
 788         inp->inp_phd = phd;
 789         LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
 790 }
 791
 792 static __inline void
 793 tcp_pcbport_remove(struct inpcb *inp)
 794 {
 795         if (inp->inp_porthash == NULL && inp->inp_phd != NULL) {
 796                 /*
 797                  * On listen tcpcb per-cpu port cache.
 798                  */
 799                 LIST_REMOVE(inp, inp_portlist);
 800                 inp->inp_phd = NULL;
 801                 /* NOTE: Don't whack inp_lport, which may be used later */
 802         }
 803 }
 804
 805 static __inline void
 806 _TCP_STATE_INC(const struct tcpcb *tp)
 807 {
 808         tcpstate_count[mycpuid].tcps_count[tp->t_state]++;
 809 }
 810
 811 static __inline void
 812 _TCP_STATE_DEC(const struct tcpcb *tp)
 813 {
 814         tcpstate_count[mycpuid].tcps_count[tp->t_state]--;
 815 }
 816
 817 static __inline void
 818 _TCP_STATE_SET(struct tcpcb *tp, int state)
 819 {
 820         tp->t_state = state;
 821         _TCP_STATE_INC(tp);
 822 }
 823
 824 static __inline void
 825 TCP_STATE_INIT(struct tcpcb *tp)
 826 {
 827         _TCP_STATE_SET(tp, TCPS_CLOSED);
 828 }
 829
 830 static __inline void
 831 TCP_STATE_TERM(struct tcpcb *tp)
 832 {
 833         KASSERT(tp->t_state != TCPS_TERMINATING, ("tcpcb was terminated"));
 834         _TCP_STATE_DEC(tp);
 835         tp->t_state = TCPS_TERMINATING;
 836 }
 837
 838 static __inline void
 839 TCP_STATE_CHANGE(struct tcpcb *tp, int state)
 840 {
 841         _TCP_STATE_DEC(tp);
 842         _TCP_STATE_SET(tp, state);
 843 }
 844
 845 static __inline void
 846 TCP_STATE_MIGRATE_START(const struct tcpcb *tp)
 847 {
 848         _TCP_STATE_DEC(tp);
 849 }
 850
 851 static __inline void
 852 TCP_STATE_MIGRATE_END(const struct tcpcb *tp)
 853 {
 854         _TCP_STATE_INC(tp);
 855 }
 856
 857 #endif /* _KERNEL */
 858
 859 #endif /* _NETINET_TCP_VAR_H_ */