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_rxtthresh_max;
  82 extern int tcp_do_sack;
  83 extern int tcp_do_smartsack;
  84 extern int tcp_do_rescuesack;
  85 extern int tcp_aggressive_rescuesack;
  86 extern int tcp_do_rfc6675;
  87 extern int tcp_rfc6675_rxt;
  88 extern int tcp_aggregate_acks;
  89 extern int tcp_eifel_rtoinc;
  90 extern int tcp_prio_synack;
  91
  92 /* TCP segment queue entry */
  93 struct tseg_qent {
  94         TAILQ_ENTRY(tseg_qent) tqe_q;
  95         int     tqe_len;                /* TCP segment data length */
  96         struct  tcphdr *tqe_th;         /* a pointer to tcp header */
  97         struct  mbuf    *tqe_m;         /* mbuf contains packet */
  98 };
  99 TAILQ_HEAD(tsegqe_head, tseg_qent);
 100 extern int      tcp_reass_maxseg;
 101 extern int      tcp_reass_qsize;
 102 #ifdef MALLOC_DECLARE
 103 MALLOC_DECLARE(M_TSEGQ);
 104 #endif
 105
 106 struct tcptemp {
 107         u_char  tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
 108         struct  tcphdr tt_t;
 109 };
 110
 111 #define tcp6cb          tcpcb  /* for KAME src sync over BSD*'s */
 112
 113 struct raw_sackblock {                          /* covers [start, end) */
 114         tcp_seq rblk_start;
 115         tcp_seq rblk_end;
 116 };
 117
 118 /* maximum number of SACK blocks that will fit in the TCP option space */
 119 #define MAX_SACK_REPORT_BLOCKS  4
 120
 121 TAILQ_HEAD(sackblock_list, sackblock);
 122
 123 struct scoreboard {
 124         int nblocks;
 125         struct sackblock_list sackblocks;
 126         tcp_seq lostseq;                        /* passed SACK lost test */
 127         struct sackblock *lastfound;            /* search hint */
 128         struct sackblock *freecache;            /* one slot free block cache */
 129 };
 130
 131 struct netmsg_tcp_timer;
 132 struct netmsg_base;
 133
 134 struct tcp_pcbport {
 135         struct inpcbport        t_phd;
 136 } __cachealign;
 137
 138 /*
 139  * Tcp control block, one per tcp; fields:
 140  * Organized for 16 byte cacheline efficiency.
 141  */
 142 struct tcpcb {
 143         struct  tsegqe_head t_segq;
 144         int     t_dupacks;              /* consecutive dup acks recd */
 145         int     t_rxtthresh;            /* # dup acks to start fast rxt */
 146         int     tt_cpu;                 /* sanity check the cpu */
 147
 148         struct  tcp_pcbport *t_pcbport; /* per-cpu local port cache for
 149                                          * accept(2)'ed sockets */
 150
 151         struct  tcp_callout *tt_rexmt;  /* retransmit timer */
 152         struct  tcp_callout *tt_persist;/* retransmit persistence */
 153         struct  tcp_callout *tt_keep;   /* keepalive */
 154         struct  tcp_callout *tt_2msl;   /* 2*msl TIME_WAIT timer */
 155         struct  tcp_callout *tt_delack; /* delayed ACK timer */
 156         struct  netmsg_tcp_timer *tt_msg; /* timer message */
 157
 158         struct  netmsg_base *tt_sndmore;/* send more segments (fairsend) */
 159
 160         struct  inpcb *t_inpcb;         /* back pointer to internet pcb */
 161         int     t_state;                /* state of this connection */
 162         u_int   t_flags;
 163 #define TF_ACKNOW       0x00000001      /* ack peer immediately */
 164 #define TF_DELACK       0x00000002      /* ack, but try to delay it */
 165 #define TF_NODELAY      0x00000004      /* don't delay packets to coalesce */
 166 #define TF_NOOPT        0x00000008      /* don't use tcp options */
 167 #define TF_SENTFIN      0x00000010      /* have sent FIN */
 168 #define TF_REQ_SCALE    0x00000020      /* have/will request window scaling */
 169 #define TF_RCVD_SCALE   0x00000040      /* other side has requested scaling */
 170 #define TF_REQ_TSTMP    0x00000080      /* have/will request timestamps */
 171 #define TF_RCVD_TSTMP   0x00000100      /* a timestamp was received in SYN */
 172 #define TF_SACK_PERMITTED 0x00000200    /* other side said I could SACK */
 173 #define TF_NEEDSYN      0x00000400      /* send SYN (implicit state) */
 174 #define TF_NEEDFIN      0x00000800      /* send FIN (implicit state) */
 175 #define TF_NOPUSH       0x00001000      /* don't push */
 176 #define TF_LISTEN       0x00002000      /* listen(2) has been called */
 177 #define TF_SIGNATURE    0x00004000      /* require MD5 digests (RFC2385) */
 178 #define TF_NCR          0x00008000      /* Non-Congestion Robustness RFC4653 */
 179 #define TF_MORETOCOME   0x00010000      /* More data to be appended to sock */
 180 #define TF_SAWFIN       0x00020000      /* FIN has been seen */
 181 #define TF_LASTIDLE     0x00040000      /* connection was previously idle */
 182 #define TF_RXWIN0SENT   0x00080000      /* sent a receiver win 0 in response */
 183 #define TF_FASTRECOVERY 0x00100000      /* in Fast Recovery */
 184 #define TF_QUEDFIN      0x00200000      /* FIN has been received */
 185 #define TF_XMITNOW      0x00400000      /* Temporarily override Nagle */
 186 #define TF_UNUSED008    0x00800000
 187 #define TF_UNUSED009    0x01000000
 188 #define TF_FORCE        0x02000000      /* Set if forcing out a byte */
 189 #define TF_ONOUTPUTQ    0x04000000      /* on t_outputq list */
 190 #define TF_FAIRSEND     0x08000000      /* fairsend is requested */
 191 #define TF_UNUSED003    0x10000000
 192 #define TF_UNUSED004    0x20000000
 193 #define TF_KEEPALIVE    0x40000000      /* temporary keepalive */
 194 #define TF_RXRESIZED    0x80000000      /* rcvbuf was resized */
 195         tcp_seq snd_up;                 /* send urgent pointer */
 196         u_long  snd_last;               /* time last data were sent */
 197
 198         tcp_seq snd_una;                /* send unacknowledged */
 199         tcp_seq snd_recover;            /* for use with Fast Recovery */
 200         tcp_seq snd_max;                /* highest sequence number sent;
 201                                          * used to recognize retransmits */
 202         tcp_seq snd_nxt;                /* send next */
 203
 204         tcp_seq snd_wl1;                /* window update seg seq number */
 205         tcp_seq snd_wl2;                /* window update seg ack number */
 206         tcp_seq iss;                    /* initial send sequence number */
 207         tcp_seq irs;                    /* initial receive sequence number */
 208
 209         tcp_seq rcv_nxt;                /* receive next */
 210         tcp_seq rcv_adv;                /* advertised window */
 211         u_long  rcv_wnd;                /* receive window */
 212         tcp_seq rcv_up;                 /* receive urgent pointer */
 213
 214         u_long  snd_wnd;                /* send window */
 215         u_long  snd_cwnd;               /* congestion-controlled window */
 216         u_long  snd_wacked;             /* bytes acked in one send window */
 217         u_long  snd_ssthresh;           /* snd_cwnd size threshold for
 218                                          * for slow start exponential to
 219                                          * linear switch */
 220
 221         int     t_rxtcur;               /* current retransmit value (ticks) */
 222         u_int   t_maxseg;               /* maximum segment size */
 223         int     t_srtt;                 /* smoothed round-trip time */
 224         int     t_rttvar;               /* variance in round-trip time */
 225
 226         u_int   t_maxopd;               /* mss plus options */
 227
 228         u_long  t_rcvtime;              /* reception inactivity time */
 229         u_long  t_starttime;            /* time connection was established */
 230         int     t_rtttime;              /* round trip time */
 231         tcp_seq t_rtseq;                /* sequence number being timed */
 232
 233         int     t_rxtshift;             /* log(2) of rexmt exp. backoff */
 234         u_int   t_rttmin;               /* minimum rtt allowed */
 235         u_int   t_rttbest;              /* best rtt we've seen */
 236         u_long  t_rttupdated;           /* number of times rtt sampled */
 237         u_long  max_sndwnd;             /* largest window peer has offered */
 238
 239         int     t_softerror;            /* possible error not yet reported */
 240 /* out-of-band data */
 241         char    t_oobflags;             /* have some */
 242         char    t_iobc;                 /* input character */
 243 #define TCPOOB_HAVEDATA 0x01
 244 #define TCPOOB_HADDATA  0x02
 245
 246 /* RFC 1323 variables */
 247         u_char  snd_scale;              /* window scaling for send window */
 248         u_char  rcv_scale;              /* window scaling for recv window */
 249         u_char  request_r_scale;        /* pending window scaling */
 250         u_long  ts_recent;              /* timestamp echo data */
 251
 252         u_long  ts_recent_age;          /* when last updated */
 253         tcp_seq last_ack_sent;
 254
 255 /* experimental */
 256         u_int   rxt_flags;
 257 #define TRXT_F_REBASERTO        0x0001  /* Recalculate RTO based on new RTT */
 258 #define TRXT_F_WASFRECOVERY     0x0002  /* was in Fast Recovery */
 259 #define TRXT_F_FIRSTACCACK      0x0004  /* Look for 1st acceptable ACK. */
 260 #define TRXT_F_FASTREXMT        0x0008  /* Did Fast Retransmit. */
 261 #define TRXT_F_EARLYREXMT       0x0010  /* Did Early (Fast) Retransmit. */
 262         int     t_srtt_prev;            /* adjusted SRTT prior to retransmit */
 263         int     t_rttvar_prev;          /* RTTVAR prior to retransmit */
 264         int     t_rxtcur_prev;          /* rexmt timeout prior to retransmit */
 265         tcp_seq snd_max_prev;           /* SND_MAX prior to retransmit */
 266         u_long  snd_cwnd_prev;          /* cwnd prior to retransmit */
 267         u_long  snd_wacked_prev;        /* prior bytes acked in send window */
 268         u_long  snd_ssthresh_prev;      /* ssthresh prior to retransmit */
 269         tcp_seq snd_recover_prev;       /* snd_recover prior to retransmit */
 270         u_long  t_badrxtwin;            /* window for retransmit recovery */
 271         u_long  t_rexmtTS;              /* timestamp of last retransmit */
 272         u_char  snd_limited;            /* segments limited transmitted */
 273
 274         u_int   sack_flags;
 275 #define TSACK_F_SACKRESCUED     0x0001  /* sent rescue SACK recovery data */
 276 #define TSACK_F_DUPSEG          0x0002  /* last seg a duplicate */
 277 #define TSACK_F_ENCLOSESEG      0x0004  /* enclosing SACK block */
 278 #define TSACK_F_SACKLEFT        0x0008  /* send SACK blocks from left side */
 279         tcp_seq rexmt_high;             /* highest seq # retransmitted + 1 */
 280         tcp_seq rexmt_rescue;           /* rescue SACKED sequence number */
 281         tcp_seq snd_max_rexmt;          /* snd_max when rexmting snd_una */
 282         struct scoreboard scb;          /* sack scoreboard */
 283         struct raw_sackblock reportblk; /* incoming segment or D-SACK block */
 284         struct raw_sackblock encloseblk;
 285         int     nsackhistory;
 286         struct raw_sackblock sackhistory[MAX_SACK_REPORT_BLOCKS]; /* reported */
 287
 288         TAILQ_ENTRY(tcpcb) t_outputq;   /* tcp_output needed list */
 289
 290         /* bandwith limitation */
 291         u_long  snd_bandwidth;          /* calculated bandwidth or 0 */
 292         u_long  snd_bwnd;               /* bandwidth-controlled window */
 293         int     t_bw_rtttime;           /* used for bandwidth calculation */
 294         tcp_seq t_bw_rtseq;             /* used for bandwidth calculation */
 295
 296 /* anti DoS counters */
 297         u_long  rcv_second;             /* start of interval second */
 298         u_long  rcv_pps;                /* received packets per second */
 299         u_long  rcv_byps;               /* received bytes per second */
 300
 301         u_int32_t       rfbuf_ts;       /* recv buffer autoscaling timestamp */
 302         int     rfbuf_cnt;              /* recv buffer autoscaling byte count */
 303
 304         int     t_keepinit;             /* time to establish connection */
 305
 306         int     t_keepidle;             /* time before keepalive probes begin */
 307         int     t_keepintvl;            /* time between keepalive probes */
 308         int     t_keepcnt;              /* maximum number of keepalive probes */
 309         int     t_maxidle;              /* time to drop after starting probes */
 310
 311         int     t_rxtsyn;               /* time spent in SYN or SYN|ACK rexmt */
 312 };
 313
 314 #define IN_FASTRECOVERY(tp)     (tp->t_flags & TF_FASTRECOVERY)
 315 #define ENTER_FASTRECOVERY(tp)  tp->t_flags |= TF_FASTRECOVERY
 316 #define EXIT_FASTRECOVERY(tp)   tp->t_flags &= ~TF_FASTRECOVERY
 317
 318 #ifdef TCP_SIGNATURE
 319 /*
 320  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
 321  * for SADB verification and lookup.
 322  */
 323 #define TCP_SIGLEN      16      /* length of computed digest in bytes */
 324 #define TCP_KEYLEN_MIN  1       /* minimum length of TCP-MD5 key */
 325 #define TCP_KEYLEN_MAX  80      /* maximum length of TCP-MD5 key */
 326 /*
 327  * Only a single SA per host may be specified at this time. An SPI is
 328  * needed in order for the KEY_ALLOCSA() lookup to work.
 329  */
 330 #define TCP_SIG_SPI     0x1000
 331 #endif /* TCP_SIGNATURE */
 332
 333 /*
 334  * TCP statistics.
 335  *
 336  * NOTE: Make sure this struct's size is multiple cache line size.
 337  */
 338 struct tcp_stats {
 339         u_long  tcps_connattempt;       /* connections initiated */
 340         u_long  tcps_accepts;           /* connections accepted */
 341         u_long  tcps_connects;          /* connections established */
 342         u_long  tcps_drops;             /* connections dropped */
 343         u_long  tcps_conndrops;         /* embryonic connections dropped */
 344         u_long  tcps_minmssdrops;       /* average minmss too low drops */
 345         u_long  tcps_closed;            /* conn. closed (includes drops) */
 346         u_long  tcps_segstimed;         /* segs where we tried to get rtt */
 347         u_long  tcps_rttupdated;        /* times we succeeded */
 348         u_long  tcps_delack;            /* delayed acks sent */
 349         u_long  tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
 350         u_long  tcps_rexmttimeo;        /* retransmit timeouts */
 351         u_long  tcps_persisttimeo;      /* persist timeouts */
 352         u_long  tcps_keeptimeo;         /* keepalive timeouts */
 353         u_long  tcps_keepprobe;         /* keepalive probes sent */
 354         u_long  tcps_keepdrops;         /* connections dropped in keepalive */
 355
 356         u_long  tcps_sndtotal;          /* total packets sent */
 357         u_long  tcps_sndpack;           /* data packets sent */
 358         u_long  tcps_sndbyte;           /* data bytes sent */
 359         u_long  tcps_sndrexmitpack;     /* data packets retransmitted */
 360         u_long  tcps_sndrexmitbyte;     /* data bytes retransmitted */
 361         u_long  tcps_sndsackrtopack;    /* packets sent by SACK after RTO */
 362         u_long  tcps_sndsackrtobyte;    /* bytes sent by SACK after RTO */
 363         u_long  tcps_sndfastrexmit;     /* Fast Retransmissions */
 364         u_long  tcps_sndearlyrexmit;    /* early Fast Retransmissions */
 365         u_long  tcps_sndlimited;        /* Limited Transmit packets */
 366         u_long  tcps_sndrtobad;         /* spurious RTO retransmissions */
 367         u_long  tcps_sndfastrexmitbad;  /* spurious Fast Retransmissions */
 368         u_long  tcps_sndearlyrexmitbad; /* spurious early Fast Retransmissions,
 369                                            a subset of tcps_sndfastrexmitbad */
 370         u_long  tcps_eifeldetected;     /* Eifel-detected spurious rexmits */
 371         u_long  tcps_rttcantdetect;     /* Eifel but not 1/2 RTT-detectable */
 372         u_long  tcps_rttdetected;       /* RTT-detected spurious RTO rexmits */
 373         u_long  tcps_sndacks;           /* ack-only packets sent */
 374         u_long  tcps_sndprobe;          /* window probes sent */
 375         u_long  tcps_sndurg;            /* packets sent with URG only */
 376         u_long  tcps_sndwinup;          /* window update-only packets sent */
 377         u_long  tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
 378         u_long  tcps_sndsackpack;       /* packets sent by SACK recovery alg */
 379         u_long  tcps_sndsackbyte;       /* bytes sent by SACK recovery */
 380         u_long  tcps_snduna3;           /* re-retransmit snd_una on 3 new seg */
 381         u_long  tcps_snduna1;           /* re-retransmit snd_una on 1 new seg */
 382         u_long  tcps_sndsackopt;        /* SACK options sent */
 383         u_long  tcps_snddsackopt;       /* D-SACK options sent */
 384         u_long  tcps_sndidle;           /* sending idle detected */
 385         u_long  tcps_sackrescue;        /* SACK rescue data packets sent */
 386         u_long  tcps_sackrescue_try;    /* SACK rescues attempted */
 387         u_long  tcps_eifelresponse;     /* Eifel responses */
 388
 389         u_long  tcps_rcvtotal;          /* total packets received */
 390         u_long  tcps_rcvpack;           /* packets received in sequence */
 391         u_long  tcps_rcvbyte;           /* bytes received in sequence */
 392         u_long  tcps_rcvbadsum;         /* packets received with ccksum errs */
 393         u_long  tcps_rcvbadoff;         /* packets received with bad offset */
 394         u_long  tcps_rcvmemdrop;        /* packets dropped for lack of memory */
 395         u_long  tcps_rcvshort;          /* packets received too short */
 396         u_long  tcps_rcvduppack;        /* duplicate-only packets received */
 397         u_long  tcps_rcvdupbyte;        /* duplicate-only bytes received */
 398         u_long  tcps_rcvpartduppack;    /* packets with some duplicate data */
 399         u_long  tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
 400         u_long  tcps_rcvoopack;         /* out-of-order packets received */
 401         u_long  tcps_rcvoobyte;         /* out-of-order bytes received */
 402         u_long  tcps_rcvpackafterwin;   /* packets with data after window */
 403         u_long  tcps_rcvbyteafterwin;   /* bytes rcvd after window */
 404         u_long  tcps_rcvafterclose;     /* packets rcvd after "close" */
 405         u_long  tcps_rcvwinprobe;       /* rcvd window probe packets */
 406         u_long  tcps_rcvdupack;         /* rcvd duplicate acks */
 407         u_long  tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
 408         u_long  tcps_rcvackpack;        /* rcvd ack packets */
 409         u_long  tcps_rcvackbyte;        /* bytes acked by rcvd acks */
 410         u_long  tcps_rcvwinupd;         /* rcvd window update packets */
 411         u_long  tcps_pawsdrop;          /* segments dropped due to PAWS */
 412         u_long  tcps_pawsaccept;        /* segments accepted, PAWS tolerance */
 413         u_long  tcps_predack;           /* times hdr predict ok for acks */
 414         u_long  tcps_preddat;           /* times hdr predict ok for data pkts */
 415         u_long  tcps_pcbcachemiss;
 416         u_long  tcps_cachedrtt;         /* times cached RTT in route updated */
 417         u_long  tcps_cachedrttvar;      /* times cached rttvar updated */
 418         u_long  tcps_cachedssthresh;    /* times cached ssthresh updated */
 419         u_long  tcps_usedrtt;           /* times RTT initialized from route */
 420         u_long  tcps_usedrttvar;        /* times RTTVAR initialized from rt */
 421         u_long  tcps_usedssthresh;      /* times ssthresh initialized from rt*/
 422         u_long  tcps_persistdrop;       /* timeout in persist state */
 423         u_long  tcps_badsyn;            /* bogus SYN, e.g. premature ACK */
 424         u_long  tcps_mturesent;         /* resends due to MTU discovery */
 425         u_long  tcps_listendrop;        /* listen queue overflows */
 426         u_long  tcps_rcvbadsackopt;     /* rcvd illegal SACK options */
 427         u_long  tcps_sackrenege;        /* times other side reneged */
 428
 429         u_long  tcps_sacksbupdate;      /* times SACK scoreboard updated */
 430         u_long  tcps_sacksboverflow;    /* times SACK scoreboard overflowed */
 431         u_long  tcps_sacksbreused;      /* times SACK sb-block reused */
 432         u_long  tcps_sacksbfailed;      /* times SACK sb update failed */
 433         u_long  tcps_sacksbfast;        /* times SACK sb-block uses cache */
 434
 435         u_long  tcps_sc_added;          /* entry added to syncache */
 436         u_long  tcps_sc_retransmitted;  /* syncache entry was retransmitted */
 437         u_long  tcps_sc_dupsyn;         /* duplicate SYN packet */
 438         u_long  tcps_sc_dropped;        /* could not reply to packet */
 439         u_long  tcps_sc_completed;      /* successful extraction of entry */
 440         u_long  tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
 441         u_long  tcps_sc_cacheoverflow;  /* syncache cache limit hit */
 442         u_long  tcps_sc_reset;          /* RST removed entry from syncache */
 443         u_long  tcps_sc_stale;          /* timed out or listen socket gone */
 444         u_long  tcps_sc_aborted;        /* syncache entry aborted */
 445         u_long  tcps_sc_badack;         /* removed due to bad ACK */
 446         u_long  tcps_sc_unreach;        /* ICMP unreachable received */
 447         u_long  tcps_sc_zonefail;       /* zalloc() failed */
 448         u_long  tcps_sc_sendcookie;     /* SYN cookie sent */
 449         u_long  tcps_sc_recvcookie;     /* SYN cookie received */
 450
 451         u_long  tcps_pad[6];            /* pad to cache line size (64B) */
 452 };
 453
 454 #ifdef _KERNEL
 455
 456 #ifndef _NETINET_TCP_FSM_H_
 457 #include <netinet/tcp_fsm.h>
 458 #endif
 459
 460 struct tcp_state_count {
 461         u_long  tcps_count[TCP_NSTATES];
 462 } __cachealign;
 463
 464 #define tcpstat tcpstats_percpu[mycpuid]
 465
 466 struct sockopt;
 467
 468 extern struct tcp_stats         tcpstats_percpu[MAXCPU];
 469 extern struct tcp_state_count   tcpstate_count[MAXCPU];
 470
 471 static const int tcprexmtthresh = 3;
 472 #endif
 473
 474 /*
 475  * Structure to hold TCP options that are only used during segment
 476  * processing (in tcp_input), but not held in the tcpcb.
 477  * It's basically used to reduce the number of parameters
 478  * to tcp_dooptions.
 479  */
 480 struct tcpopt {
 481         u_int           to_flags;       /* which options are present */
 482 #define TOF_TS                  0x0001  /* timestamp */
 483 #define TOF_MSS                 0x0010
 484 #define TOF_SCALE               0x0020
 485 #define TOF_SACK_PERMITTED      0x0040
 486 #define TOF_SACK                0x0080
 487 #define TOF_SIGNATURE           0x0100  /* signature option present */
 488 #define TOF_SIGLEN              0x0200  /* signature length valid (RFC2385) */
 489 #define TOF_DSACK               0x0400  /* D-SACK */
 490 #define TOF_SACK_REDUNDANT      0x0800  /* all SACK blocks are known */
 491         u_int32_t       to_tsval;
 492         u_int32_t       to_tsecr;
 493         u_int16_t       to_mss;
 494         u_int8_t        to_requested_s_scale;
 495         u_int8_t        to_nsackblocks;
 496         struct raw_sackblock *to_sackblocks;
 497 };
 498
 499 struct syncache {
 500         inp_gen_t       sc_inp_gencnt;          /* pointer check */
 501         struct          tcpcb *sc_tp;           /* tcb for listening socket */
 502         struct          mbuf *sc_ipopts;        /* source route */
 503         struct          in_conninfo sc_inc;     /* addresses */
 504 #define sc_route        sc_inc.inc_route
 505 #define sc_route6       sc_inc.inc6_route
 506         u_int32_t       sc_tsrecent;
 507         tcp_seq         sc_irs;                 /* seq from peer */
 508         tcp_seq         sc_iss;                 /* our ISS */
 509         u_long          sc_rxttime;             /* retransmit time */
 510         u_int16_t       sc_rxtslot;             /* retransmit counter */
 511         u_int16_t       sc_peer_mss;            /* peer's MSS */
 512         u_int16_t       sc_wnd;                 /* advertised window */
 513         u_int8_t        sc_requested_s_scale:4,
 514                         sc_request_r_scale:4;
 515         u_int8_t        sc_flags;
 516 #define SCF_NOOPT               0x01            /* no TCP options */
 517 #define SCF_WINSCALE            0x02            /* negotiated window scaling */
 518 #define SCF_TIMESTAMP           0x04            /* negotiated timestamps */
 519 #define SCF_UNUSED              0x08            /* unused */
 520 #define SCF_UNREACH             0x10            /* icmp unreachable received */
 521 #define SCF_SACK_PERMITTED      0x20            /* saw SACK permitted option */
 522 #define SCF_SIGNATURE           0x40            /* send MD5 digests */
 523 #define SCF_MARKER              0x80            /* not a real entry */
 524         int             sc_rxtused;             /* time spent in SYN|ACK rxt */
 525         u_long          sc_sndwnd;              /* send window */
 526         TAILQ_ENTRY(syncache) sc_hash;
 527         TAILQ_ENTRY(syncache) sc_timerq;
 528 };
 529
 530 struct syncache_head {
 531         TAILQ_HEAD(, syncache)  sch_bucket;
 532         u_int           sch_length;
 533 };
 534
 535 #define intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
 536 #define sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
 537
 538 /*
 539  * The smoothed round-trip time and estimated variance
 540  * are stored as fixed point numbers scaled by the values below.
 541  * For convenience, these scales are also used in smoothing the average
 542  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
 543  * With these scales, srtt has 3 bits to the right of the binary point,
 544  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
 545  * binary point, and is smoothed with an ALPHA of 0.75.
 546  */
 547 #define TCP_RTT_SCALE           32      /* multiplier for srtt; 3 bits frac. */
 548 #define TCP_RTT_SHIFT           5       /* shift for srtt; 3 bits frac. */
 549 #define TCP_RTTVAR_SCALE        16      /* multiplier for rttvar; 2 bits */
 550 #define TCP_RTTVAR_SHIFT        4       /* shift for rttvar; 2 bits */
 551 #define TCP_DELTA_SHIFT         2       /* see tcp_input.c */
 552
 553 /*
 554  * The initial retransmission should happen at rtt + 4 * rttvar.
 555  * Because of the way we do the smoothing, srtt and rttvar
 556  * will each average +1/2 tick of bias.  When we compute
 557  * the retransmit timer, we want 1/2 tick of rounding and
 558  * 1 extra tick because of +-1/2 tick uncertainty in the
 559  * firing of the timer.  The bias will give us exactly the
 560  * 1.5 tick we need.  But, because the bias is
 561  * statistical, we have to test that we don't drop below
 562  * the minimum feasible timer (which is 2 ticks).
 563  * This version of the macro adapted from a paper by Lawrence
 564  * Brakmo and Larry Peterson which outlines a problem caused
 565  * by insufficient precision in the original implementation,
 566  * which results in inappropriately large RTO values for very
 567  * fast networks.
 568  */
 569 #define TCP_REXMTVAL(tp) \
 570         max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
 571           + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
 572
 573 /*
 574  * TCB structure exported to user-land via sysctl(3).
 575  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
 576  * included.  Not all of our clients do.
 577  */
 578 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
 579 struct  xtcpcb {
 580         size_t  xt_len;
 581         struct  inpcb   xt_inp;
 582         struct  tcpcb   xt_tp;
 583         struct  xsocket xt_socket;
 584         u_quad_t        xt_alignment_hack;
 585 };
 586 #endif
 587
 588 /*
 589  * Names for TCP sysctl objects
 590  */
 591 #define TCPCTL_DO_RFC1323       1       /* use RFC-1323 extensions */
 592 /* 2 was TCPCTL_DO_RFC1644 */
 593 #define TCPCTL_MSSDFLT          3       /* MSS default */
 594 #define TCPCTL_STATS            4       /* statistics (read-only) */
 595 #define TCPCTL_RTTDFLT          5       /* default RTT estimate */
 596 #define TCPCTL_KEEPIDLE         6       /* keepalive idle timer */
 597 #define TCPCTL_KEEPINTVL        7       /* interval to send keepalives */
 598 #define TCPCTL_SENDSPACE        8       /* send buffer space */
 599 #define TCPCTL_RECVSPACE        9       /* receive buffer space */
 600 #define TCPCTL_KEEPINIT         10      /* timeout for establishing syn */
 601 #define TCPCTL_PCBLIST          11      /* list of all outstanding PCBs */
 602 #define TCPCTL_DELACKTIME       12      /* time before sending delayed ACK */
 603 #define TCPCTL_V6MSSDFLT        13      /* MSS default for IPv6 */
 604 #define TCPCTL_MAXID            14
 605
 606 #define TCPCTL_NAMES { \
 607         { 0, 0 }, \
 608         { "rfc1323", CTLTYPE_INT }, \
 609           { "reserved", CTLTYPE_INT},   /* was rfc1644 */       \
 610         { "mssdflt", CTLTYPE_INT }, \
 611         { "stats", CTLTYPE_STRUCT }, \
 612         { "rttdflt", CTLTYPE_INT }, \
 613         { "keepidle", CTLTYPE_INT }, \
 614         { "keepintvl", CTLTYPE_INT }, \
 615         { "sendspace", CTLTYPE_INT }, \
 616         { "recvspace", CTLTYPE_INT }, \
 617         { "keepinit", CTLTYPE_INT }, \
 618         { "pcblist", CTLTYPE_STRUCT }, \
 619         { "delacktime", CTLTYPE_INT }, \
 620         { "v6mssdflt", CTLTYPE_INT }, \
 621 }
 622
 623 #ifdef _KERNEL
 624 #ifdef SYSCTL_DECL
 625 SYSCTL_DECL(_net_inet_tcp);
 626 #endif
 627
 628 #define TCP_DO_SACK(tp)         ((tp)->t_flags & TF_SACK_PERMITTED)
 629 #define TCP_DO_NCR(tp)          (((tp)->t_flags & TF_NCR) && TCP_DO_SACK((tp)))
 630 #define TCP_SACK_BLKEND(len, thflags) \
 631         ((len) + (((thflags) & TH_FIN) != 0))
 632
 633 TAILQ_HEAD(tcpcbackqhead,tcpcb);
 634
 635 extern  struct inpcbinfo tcbinfo[];
 636 extern  struct tcpcbackqhead tcpcbackq[];
 637
 638 extern  int tcp_mssdflt;        /* XXX */
 639 extern  int tcp_minmss;
 640 extern  int tcp_delack_enabled;
 641 extern  int path_mtu_discovery;
 642
 643 struct ip;
 644 union netmsg;
 645
 646 int      tcp_addrcpu(in_addr_t faddr, in_port_t fport,
 647             in_addr_t laddr, in_port_t lport);
 648 struct lwkt_port *
 649         tcp_addrport(in_addr_t faddr, in_port_t fport,
 650             in_addr_t laddr, in_port_t lport);
 651 struct lwkt_port *tcp_addrport0(void);
 652 void     tcp_canceltimers (struct tcpcb *);
 653 struct tcpcb *
 654          tcp_close (struct tcpcb *);
 655 void     tcp_ctlinput(union netmsg *);
 656 void     tcp_ctloutput(union netmsg *);
 657 struct netmsg_pr_ctloutput *tcp_ctloutmsg(struct sockopt *);
 658 inp_notify_t tcp_get_inpnotify(int cmd, const struct sockaddr *sa,
 659             int *arg, struct ip **ip0, int *cpuid);
 660 struct tcpcb *
 661          tcp_drop (struct tcpcb *, int);
 662 void     tcp_drain (void);
 663 void     tcp_init (void);
 664 void     tcp_thread_init (void);
 665 int      tcp_input (struct mbuf **, int *, int);
 666 void     tcp_mss (struct tcpcb *, int);
 667 int      tcp_mssopt (struct tcpcb *);
 668 void     tcp_drop_syn_sent (struct inpcb *, int);
 669 void     tcp_mtudisc (struct inpcb *, int);
 670 void     tcp_newtcpcb (struct inpcb *);
 671 int      tcp_output(struct tcpcb *);
 672 int      tcp_output_fair(struct tcpcb *);
 673 void     tcp_output_init(struct tcpcb *);
 674 void     tcp_output_cancel(struct tcpcb *);
 675 boolean_t
 676          tcp_output_pending(struct tcpcb *);
 677 void     tcp_quench (struct inpcb *, int);
 678 void     tcp_respond (struct tcpcb *, void *,
 679             struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
 680 struct rtentry *
 681          tcp_rtlookup (struct in_conninfo *);
 682 int      tcp_sack_bytes_below(const struct scoreboard *scb, tcp_seq seq);
 683 void     tcp_sack_destroy(struct scoreboard *scb);
 684 void     tcp_sack_discard(struct tcpcb *tp);
 685 void     tcp_sack_report_cleanup(struct tcpcb *tp);
 686 boolean_t
 687          tcp_sack_report_needed(const struct tcpcb *tp);
 688 int      tcp_sack_ndsack_blocks(const struct raw_sackblock *blocks,
 689             const int numblocks, tcp_seq snd_una);
 690 void     tcp_sack_fill_report(struct tcpcb *tp, u_char *opt, u_int *plen);
 691 boolean_t
 692          tcp_sack_has_sacked(const struct scoreboard *scb, u_int amount);
 693 void     tcp_sack_tcpcb_init(struct tcpcb *tp);
 694 uint32_t tcp_sack_compute_pipe(const struct tcpcb *tp);
 695 boolean_t
 696          tcp_sack_nextseg(struct tcpcb *tp, tcp_seq *nextrexmt, uint32_t *len,
 697                           boolean_t *rescue);
 698 boolean_t
 699          tcp_sack_islost(const struct scoreboard *scb, tcp_seq seq);
 700 void     tcp_sack_update_lostseq(struct scoreboard *scb, tcp_seq snd_una,
 701             u_int maxseg, int rxtthresh);
 702 #ifdef later
 703 void     tcp_sack_revert_scoreboard(struct scoreboard *scb, tcp_seq snd_una,
 704                                     u_int maxseg);
 705 void     tcp_sack_save_scoreboard(struct scoreboard *scb);
 706 #endif
 707 void     tcp_sack_skip_sacked(struct scoreboard *scb, tcp_seq *prexmt);
 708 uint32_t tcp_sack_first_unsacked_len(const struct tcpcb *tp);
 709 void     tcp_sack_update_scoreboard(struct tcpcb *tp, struct tcpopt *to);
 710 void     tcp_save_congestion_state(struct tcpcb *tp);
 711 void     tcp_revert_congestion_state(struct tcpcb *tp);
 712 void     tcp_setpersist (struct tcpcb *);
 713 struct tcptemp *tcp_maketemplate (struct tcpcb *);
 714 void     tcp_freetemplate (struct tcptemp *);
 715 void     tcp_fillheaders (struct tcpcb *, void *, void *, boolean_t);
 716 struct lwkt_port *
 717          tcp_soport(struct socket *, struct sockaddr *, struct mbuf **);
 718 struct lwkt_port *
 719          tcp_ctlport(int, struct sockaddr *, void *, int *);
 720 struct lwkt_port *
 721          tcp_initport(void);
 722 struct tcpcb *
 723          tcp_timers (struct tcpcb *, int);
 724 void     tcp_trace (short, short, struct tcpcb *, void *, struct tcphdr *,
 725                         int);
 726 void     tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
 727 u_long   tcp_initial_window(struct tcpcb *tp);
 728 void     tcp_timer_keep_activity(struct tcpcb *tp, int thflags);
 729 void     syncache_init(void);
 730 void     syncache_unreach(struct in_conninfo *, const struct tcphdr *);
 731 int      syncache_expand(struct in_conninfo *, struct tcphdr *,
 732              struct socket **, struct mbuf *);
 733 int      syncache_add(struct in_conninfo *, struct tcpopt *,
 734              struct tcphdr *, struct socket *, struct mbuf *);
 735 void     syncache_chkrst(struct in_conninfo *, struct tcphdr *);
 736 void     syncache_badack(struct in_conninfo *);
 737 void     syncache_destroy(struct tcpcb *tp, struct tcpcb *new_tp);
 738
 739 #ifdef TCP_SIGNATURE
 740 int tcpsignature_apply(void *fstate, void *data, unsigned int len);
 741 int tcpsignature_compute(struct mbuf *m, int len, int tcpoptlen,
 742                 u_char *buf, u_int direction);
 743 #endif /* TCP_SIGNATURE */
 744
 745 extern  struct pr_usrreqs tcp_usrreqs;
 746 extern  u_long tcp_sendspace;
 747 extern  u_long tcp_recvspace;
 748 tcp_seq tcp_new_isn (struct tcpcb *);
 749
 750 void    tcp_pcbport_create(struct tcpcb *);
 751 void    tcp_pcbport_destroy(struct tcpcb *);
 752 void    tcp_pcbport_merge_oncpu(struct tcpcb *);
 753
 754 static __inline void
 755 tcp_pcbport_insert(struct tcpcb *ltp, struct inpcb *inp)
 756 {
 757         struct inpcbport *phd;
 758         int cpu;
 759
 760         if (inp->inp_lport != ltp->t_inpcb->inp_lport) {
 761                 /*
 762                  * This could happen with 'ipfw forward'.
 763                  */
 764                 in_pcbinsporthash_lport(inp);
 765                 return;
 766         }
 767
 768         cpu = mycpuid;
 769         KASSERT(cpu < ncpus2, ("invalid cpu%d", cpu));
 770         phd = &ltp->t_pcbport[cpu].t_phd;
 771
 772         /*
 773          * NOTE:
 774          * Set inp_porthash NULL and set inp_phd properly,
 775          * so that tcp_pcbport_remove() could tell that this
 776          * inpcb is on the listen tcpcb per-cpu port cache.
 777          */
 778         inp->inp_porthash = NULL;
 779         inp->inp_phd = phd;
 780         LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
 781 }
 782
 783 static __inline void
 784 tcp_pcbport_remove(struct inpcb *inp)
 785 {
 786         if (inp->inp_porthash == NULL && inp->inp_phd != NULL) {
 787                 /*
 788                  * On listen tcpcb per-cpu port cache.
 789                  */
 790                 LIST_REMOVE(inp, inp_portlist);
 791                 inp->inp_phd = NULL;
 792                 /* NOTE: Don't whack inp_lport, which may be used later */
 793         }
 794 }
 795
 796 static __inline void
 797 _TCP_STATE_INC(const struct tcpcb *tp)
 798 {
 799         tcpstate_count[mycpuid].tcps_count[tp->t_state]++;
 800 }
 801
 802 static __inline void
 803 _TCP_STATE_DEC(const struct tcpcb *tp)
 804 {
 805         tcpstate_count[mycpuid].tcps_count[tp->t_state]--;
 806 }
 807
 808 static __inline void
 809 _TCP_STATE_SET(struct tcpcb *tp, int state)
 810 {
 811         tp->t_state = state;
 812         _TCP_STATE_INC(tp);
 813 }
 814
 815 static __inline void
 816 TCP_STATE_INIT(struct tcpcb *tp)
 817 {
 818         _TCP_STATE_SET(tp, TCPS_CLOSED);
 819 }
 820
 821 static __inline void
 822 TCP_STATE_TERM(struct tcpcb *tp)
 823 {
 824         KASSERT(tp->t_state != TCPS_TERMINATING, ("tcpcb was terminated"));
 825         _TCP_STATE_DEC(tp);
 826         tp->t_state = TCPS_TERMINATING;
 827 }
 828
 829 static __inline void
 830 TCP_STATE_CHANGE(struct tcpcb *tp, int state)
 831 {
 832         _TCP_STATE_DEC(tp);
 833         _TCP_STATE_SET(tp, state);
 834 }
 835
 836 static __inline void
 837 TCP_STATE_MIGRATE_START(const struct tcpcb *tp)
 838 {
 839         _TCP_STATE_DEC(tp);
 840 }
 841
 842 static __inline void
 843 TCP_STATE_MIGRATE_END(const struct tcpcb *tp)
 844 {
 845         _TCP_STATE_INC(tp);
 846 }
 847
 848 #endif /* _KERNEL */
 849
 850 #endif /* _NETINET_TCP_VAR_H_ */