workbench/network/common/include/netinet/tcp_var.h

   1 /*
   2  * Copyright (c) 1982, 1986, 1993, 1994
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  *
  33  *      @(#)tcp_var.h   8.3 (Berkeley) 4/10/94
  34  * $Id$
  35  */
  36
  37 #ifndef _NETINET_TCP_VAR_H_
  38 #define _NETINET_TCP_VAR_H_
  39 /*
  40  * Kernel variables for tcp.
  41  */
  42
  43 /*
  44  * Tcp control block, one per tcp; fields:
  45  */
  46 struct tcpcb {
  47         struct  mbuf *t_segq; /* sequencing queue */
  48         short   t_state;                /* state of this connection */
  49         short   t_timer[TCPT_NTIMERS];  /* tcp timers */
  50         short   t_rxtshift;             /* log(2) of rexmt exp. backoff */
  51         short   t_rxtcur;               /* current retransmit value */
  52         short   t_dupacks;              /* consecutive dup acks recd */
  53         u_short t_maxseg;               /* maximum segment size */
  54         u_short t_maxopd;               /* mss plus options */
  55         char    t_force;                /* 1 if forcing out a byte */
  56         u_short t_flags;
  57 #define TF_ACKNOW       0x0001          /* ack peer immediately */
  58 #define TF_DELACK       0x0002          /* ack, but try to delay it */
  59 #define TF_NODELAY      0x0004          /* don't delay packets to coalesce */
  60 #define TF_NOOPT        0x0008          /* don't use tcp options */
  61 #define TF_SENTFIN      0x0010          /* have sent FIN */
  62 #define TF_REQ_SCALE    0x0020          /* have/will request window scaling */
  63 #define TF_RCVD_SCALE   0x0040          /* other side has requested scaling */
  64 #define TF_REQ_TSTMP    0x0080          /* have/will request timestamps */
  65 #define TF_RCVD_TSTMP   0x0100          /* a timestamp was received in SYN */
  66 #define TF_SACK_PERMIT  0x0200          /* other side said I could SACK */
  67 #define TF_NEEDSYN      0x0400          /* send SYN (implicit state) */
  68 #define TF_NEEDFIN      0x0800          /* send FIN (implicit state) */
  69 #define TF_NOPUSH       0x1000          /* don't push */
  70 #define TF_REQ_CC       0x2000          /* have/will request CC */
  71 #define TF_RCVD_CC      0x4000          /* a CC was received in SYN */
  72
  73         struct  tcpiphdr *t_template;   /* skeletal packet for transmit */
  74         struct  inpcb *t_inpcb;         /* back pointer to internet pcb */
  75 /*
  76  * The following fields are used as in the protocol specification.
  77  * See RFC783, Dec. 1981, page 21.
  78  */
  79 /* send sequence variables */
  80         tcp_seq snd_una;                /* send unacknowledged */
  81         tcp_seq snd_nxt;                /* send next */
  82         tcp_seq snd_up;                 /* send urgent pointer */
  83         tcp_seq snd_wl1;                /* window update seg seq number */
  84         tcp_seq snd_wl2;                /* window update seg ack number */
  85         tcp_seq iss;                    /* initial send sequence number */
  86         u_long  snd_wnd;                /* send window */
  87 /* receive sequence variables */
  88         u_long  rcv_wnd;                /* receive window */
  89         tcp_seq rcv_nxt;                /* receive next */
  90         tcp_seq rcv_up;                 /* receive urgent pointer */
  91         tcp_seq irs;                    /* initial receive sequence number */
  92 /*
  93  * Additional variables for this implementation.
  94  */
  95 /* receive variables */
  96         tcp_seq rcv_adv;                /* advertised window */
  97 /* retransmit variables */
  98         tcp_seq snd_max;                /* highest sequence number sent;
  99                                          * used to recognize retransmits
 100                                          */
 101 /* congestion control (for slow start, source quench, retransmit after loss) */
 102         u_long  snd_cwnd;               /* congestion-controlled window */
 103         u_long  snd_ssthresh;           /* snd_cwnd size threshhold for
 104                                          * for slow start exponential to
 105                                          * linear switch
 106                                          */
 107 /*
 108  * transmit timing stuff.  See below for scale of srtt and rttvar.
 109  * "Variance" is actually smoothed difference.
 110  */
 111         short   t_idle;                 /* inactivity time */
 112         short   t_rtt;                  /* round trip time */
 113         tcp_seq t_rtseq;                /* sequence number being timed */
 114         short   t_srtt;                 /* smoothed round-trip time */
 115         short   t_rttvar;               /* variance in round-trip time */
 116         u_short t_rttmin;               /* minimum rtt allowed */
 117         u_long  max_sndwnd;             /* largest window peer has offered */
 118
 119 /* out-of-band data */
 120         char    t_oobflags;             /* have some */
 121         char    t_iobc;                 /* input character */
 122 #define TCPOOB_HAVEDATA 0x01
 123 #define TCPOOB_HADDATA  0x02
 124         short   t_softerror;            /* possible error not yet reported */
 125
 126 /* RFC 1323 variables */
 127         u_char  snd_scale;              /* window scaling for send window */
 128         u_char  rcv_scale;              /* window scaling for recv window */
 129         u_char  request_r_scale;        /* pending window scaling */
 130         u_char  requested_s_scale;
 131         u_long  ts_recent;              /* timestamp echo data */
 132         u_long  ts_recent_age;          /* when last updated */
 133         tcp_seq last_ack_sent;
 134 /* RFC 1644 variables */
 135         tcp_cc  cc_send;                /* send connection count */
 136         tcp_cc  cc_recv;                /* receive connection count */
 137         u_long  t_duration;             /* connection duration */
 138
 139 /* TUBA stuff */
 140         caddr_t t_tuba_pcb;             /* next level down pcb for TCP over z */
 141 };
 142
 143 /*
 144  * Structure to hold TCP options that are only used during segment
 145  * processing (in tcp_input), but not held in the tcpcb.
 146  * It's basically used to reduce the number of parameters
 147  * to tcp_dooptions.
 148  */
 149 struct tcpopt {
 150         u_long  to_flag;                /* which options are present */
 151 #define TOF_TS          0x0001          /* timestamp */
 152 #define TOF_CC          0x0002          /* CC and CCnew are exclusive */
 153 #define TOF_CCNEW       0x0004
 154 #define TOF_CCECHO      0x0008
 155         u_long  to_tsval;
 156         u_long  to_tsecr;
 157         tcp_cc  to_cc;          /* holds CC or CCnew */
 158         tcp_cc  to_ccecho;
 159 };
 160
 161 /*
 162  * The TAO cache entry which is stored in the protocol family specific
 163  * portion of the route metrics.
 164  */
 165 struct rmxp_tao {
 166         tcp_cc  tao_cc;                 /* latest CC in valid SYN */
 167         tcp_cc  tao_ccsent;             /* latest CC sent to peer */
 168         u_short tao_mssopt;             /* peer's cached MSS */
 169 #ifdef notyet
 170         u_short tao_flags;              /* cache status flags */
 171 #define TAOF_DONT       0x0001          /* peer doesn't understand rfc1644 */
 172 #define TAOF_OK         0x0002          /* peer does understand rfc1644 */
 173 #define TAOF_UNDEF      0               /* we don't know yet */
 174 #endif /* notyet */
 175 };
 176 #define rmx_taop(r)     ((struct rmxp_tao *)(r).rmx_filler)
 177
 178 #define intotcpcb(ip)   ((struct tcpcb *)(ip)->inp_ppcb)
 179 #define sototcpcb(so)   (intotcpcb(sotoinpcb(so)))
 180
 181 /*
 182  * The smoothed round-trip time and estimated variance
 183  * are stored as fixed point numbers scaled by the values below.
 184  * For convenience, these scales are also used in smoothing the average
 185  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
 186  * With these scales, srtt has 3 bits to the right of the binary point,
 187  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
 188  * binary point, and is smoothed with an ALPHA of 0.75.
 189  */
 190 #define TCP_RTT_SCALE           8       /* multiplier for srtt; 3 bits frac. */
 191 #define TCP_RTT_SHIFT           3       /* shift for srtt; 3 bits frac. */
 192 #define TCP_RTTVAR_SCALE        4       /* multiplier for rttvar; 2 bits */
 193 #define TCP_RTTVAR_SHIFT        2       /* shift for rttvar; 2 bits */
 194
 195 /*
 196  * The initial retransmission should happen at rtt + 4 * rttvar.
 197  * Because of the way we do the smoothing, srtt and rttvar
 198  * will each average +1/2 tick of bias.  When we compute
 199  * the retransmit timer, we want 1/2 tick of rounding and
 200  * 1 extra tick because of +-1/2 tick uncertainty in the
 201  * firing of the timer.  The bias will give us exactly the
 202  * 1.5 tick we need.  But, because the bias is
 203  * statistical, we have to test that we don't drop below
 204  * the minimum feasible timer (which is 2 ticks).
 205  * This macro assumes that the value of TCP_RTTVAR_SCALE
 206  * is the same as the multiplier for rttvar.
 207  */
 208 #define TCP_REXMTVAL(tp) \
 209         (((tp)->t_srtt >> TCP_RTT_SHIFT) + (tp)->t_rttvar)
 210
 211 /* XXX
 212  * We want to avoid doing m_pullup on incoming packets but that
 213  * means avoiding dtom on the tcp reassembly code.  That in turn means
 214  * keeping an mbuf pointer in the reassembly queue (since we might
 215  * have a cluster).  As a quick hack, the source & destination
 216  * port numbers (which are no longer needed once we've located the
 217  * tcpcb) are overlayed with an mbuf pointer.
 218  */
 219 #define REASS_MBUF(ti) (*(struct mbuf **)&((ti)->ti_t))
 220
 221 /*
 222  * TCP statistics.
 223  * Many of these should be kept per connection,
 224  * but that's inconvenient at the moment.
 225  */
 226 struct  tcpstat {
 227         u_long  tcps_connattempt;       /* connections initiated */
 228         u_long  tcps_accepts;           /* connections accepted */
 229         u_long  tcps_connects;          /* connections established */
 230         u_long  tcps_drops;             /* connections dropped */
 231         u_long  tcps_conndrops;         /* embryonic connections dropped */
 232         u_long  tcps_closed;            /* conn. closed (includes drops) */
 233         u_long  tcps_segstimed;         /* segs where we tried to get rtt */
 234         u_long  tcps_rttupdated;        /* times we succeeded */
 235         u_long  tcps_delack;            /* delayed acks sent */
 236         u_long  tcps_timeoutdrop;       /* conn. dropped in rxmt timeout */
 237         u_long  tcps_rexmttimeo;        /* retransmit timeouts */
 238         u_long  tcps_persisttimeo;      /* persist timeouts */
 239         u_long  tcps_keeptimeo;         /* keepalive timeouts */
 240         u_long  tcps_keepprobe;         /* keepalive probes sent */
 241         u_long  tcps_keepdrops;         /* connections dropped in keepalive */
 242
 243         u_long  tcps_sndtotal;          /* total packets sent */
 244         u_long  tcps_sndpack;           /* data packets sent */
 245         u_long  tcps_sndbyte;           /* data bytes sent */
 246         u_long  tcps_sndrexmitpack;     /* data packets retransmitted */
 247         u_long  tcps_sndrexmitbyte;     /* data bytes retransmitted */
 248         u_long  tcps_sndacks;           /* ack-only packets sent */
 249         u_long  tcps_sndprobe;          /* window probes sent */
 250         u_long  tcps_sndurg;            /* packets sent with URG only */
 251         u_long  tcps_sndwinup;          /* window update-only packets sent */
 252         u_long  tcps_sndctrl;           /* control (SYN|FIN|RST) packets sent */
 253
 254         u_long  tcps_rcvtotal;          /* total packets received */
 255         u_long  tcps_rcvpack;           /* packets received in sequence */
 256         u_long  tcps_rcvbyte;           /* bytes received in sequence */
 257         u_long  tcps_rcvbadsum;         /* packets received with ccksum errs */
 258         u_long  tcps_rcvbadoff;         /* packets received with bad offset */
 259         u_long  tcps_rcvshort;          /* packets received too short */
 260         u_long  tcps_rcvduppack;        /* duplicate-only packets received */
 261         u_long  tcps_rcvdupbyte;        /* duplicate-only bytes received */
 262         u_long  tcps_rcvpartduppack;    /* packets with some duplicate data */
 263         u_long  tcps_rcvpartdupbyte;    /* dup. bytes in part-dup. packets */
 264         u_long  tcps_rcvoopack;         /* out-of-order packets received */
 265         u_long  tcps_rcvoobyte;         /* out-of-order bytes received */
 266         u_long  tcps_rcvpackafterwin;   /* packets with data after window */
 267         u_long  tcps_rcvbyteafterwin;   /* bytes rcvd after window */
 268         u_long  tcps_rcvafterclose;     /* packets rcvd after "close" */
 269         u_long  tcps_rcvwinprobe;       /* rcvd window probe packets */
 270         u_long  tcps_rcvdupack;         /* rcvd duplicate acks */
 271         u_long  tcps_rcvacktoomuch;     /* rcvd acks for unsent data */
 272         u_long  tcps_rcvackpack;        /* rcvd ack packets */
 273         u_long  tcps_rcvackbyte;        /* bytes acked by rcvd acks */
 274         u_long  tcps_rcvwinupd;         /* rcvd window update packets */
 275         u_long  tcps_pawsdrop;          /* segments dropped due to PAWS */
 276         u_long  tcps_predack;           /* times hdr predict ok for acks */
 277         u_long  tcps_preddat;           /* times hdr predict ok for data pkts */
 278         u_long  tcps_pcbcachemiss;
 279 };
 280
 281 /*
 282  * Names for TCP sysctl objects
 283  */
 284 #define TCPCTL_DO_RFC1323       1       /* use RFC-1323 extensions */
 285 #define TCPCTL_DO_RFC1644       2       /* use RFC-1644 extensions */
 286 #define TCPCTL_MSSDFLT          3       /* MSS default */
 287 #define TCPCTL_STATS            4       /* statistics (read-only) */
 288 #define TCPCTL_RTTDFLT          5       /* default RTT estimate */
 289 #define TCPCTL_KEEPIDLE         6       /* keepalive idle timer */
 290 #define TCPCTL_KEEPINTVL        7       /* interval to send keepalives */
 291 #define TCPCTL_SENDSPACE        8       /* send buffer space */
 292 #define TCPCTL_RECVSPACE        9       /* receive buffer space */
 293 #define TCPCTL_MAXID            10
 294
 295 #define TCPCTL_NAMES { \
 296         { 0, 0 }, \
 297         { "rfc1323", CTLTYPE_INT }, \
 298         { "rfc1644", CTLTYPE_INT }, \
 299         { "mssdflt", CTLTYPE_INT }, \
 300         { "stats", CTLTYPE_STRUCT }, \
 301         { "rttdflt", CTLTYPE_INT }, \
 302         { "keepidle", CTLTYPE_INT }, \
 303         { "keepintvl", CTLTYPE_INT }, \
 304         { "sendspace", CTLTYPE_INT }, \
 305         { "recvspace", CTLTYPE_INT }, \
 306 }
 307
 308 #ifdef KERNEL
 309 extern  struct inpcbhead tcb;           /* head of queue of active tcpcb's */
 310 extern  struct inpcbinfo tcbinfo;
 311 extern  struct tcpstat tcpstat; /* tcp statistics */
 312 extern  int tcp_do_rfc1323;     /* XXX */
 313 extern  int tcp_do_rfc1644;     /* XXX */
 314 extern  int tcp_mssdflt;        /* XXX */
 315 extern  u_long tcp_now;         /* for RFC 1323 timestamps */
 316 extern  int tcp_rttdflt;        /* XXX */
 317 extern  u_short tcp_lastport;   /* last assigned port */
 318
 319 int      tcp_attach __P((struct socket *));
 320 void     tcp_canceltimers __P((struct tcpcb *));
 321 struct tcpcb *
 322          tcp_close __P((struct tcpcb *));
 323 int      tcp_connect __P((struct tcpcb *, struct mbuf *));
 324 void     tcp_ctlinput __P((int, struct sockaddr *, void *));
 325 int      tcp_ctloutput __P((int, struct socket *, int, int, struct mbuf **));
 326 struct tcpcb *
 327          tcp_disconnect __P((struct tcpcb *));
 328 struct tcpcb *
 329          tcp_drop __P((struct tcpcb *, int));
 330 void     tcp_dooptions __P((struct tcpcb *,
 331             u_char *, int, struct tcpiphdr *, struct tcpopt *));
 332 void     tcp_drain __P((void));
 333 void     tcp_fasttimo __P((void));
 334 struct rmxp_tao *
 335          tcp_gettaocache __P((struct inpcb *));
 336 void     tcp_init __P((void));
 337 void     tcp_input __P((void *arg, ...));
 338 void     tcp_mss __P((struct tcpcb *, int));
 339 int      tcp_mssopt __P((struct tcpcb *));
 340 struct tcpcb *
 341          tcp_newtcpcb __P((struct inpcb *));
 342 void     tcp_notify __P((struct inpcb *, int));
 343 int      tcp_output __P((struct tcpcb *));
 344 void     tcp_pulloutofband __P((struct socket *,
 345             struct tcpiphdr *, struct mbuf *));
 346 void     tcp_quench __P((struct inpcb *, int));
 347 int      tcp_reass __P((struct tcpcb *, struct tcpiphdr *, struct mbuf *));
 348 void     tcp_respond __P((struct tcpcb *,
 349             struct tcpiphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
 350 struct rtentry *
 351          tcp_rtlookup __P((struct inpcb *));
 352 void     tcp_setpersist __P((struct tcpcb *));
 353 void     tcp_slowtimo __P((void));
 354 int      tcp_sysctl __P((int *, u_int, void *, size_t *, void *, size_t));
 355 struct tcpiphdr *
 356          tcp_template __P((struct tcpcb *));
 357 struct tcpcb *
 358          tcp_timers __P((struct tcpcb *, int));
 359 void     tcp_trace __P((int, int, struct tcpcb *, struct tcpiphdr *, int));
 360 struct tcpcb *
 361          tcp_usrclosed __P((struct tcpcb *));
 362 int      tcp_usrreq __P((struct socket *,
 363             int, struct mbuf *, struct mbuf *, struct mbuf *));
 364 void     tcp_xmit_timer __P((struct tcpcb *, int));
 365 #endif /* KERNEL */
 366
 367 #endif /* _NETINET_TCP_VAR_H_ */