/* * tcp_var.h * * Derived from: * * Copyright (c) 1982, 1986 Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms are permitted * provided that this notice is preserved and that due credit is given * to the University of California at Berkeley. The name of the University * may not be used to endorse or promote products derived from this * software without specific prior written permission. This software * is provided ``as is'' without express or implied warranty. * * @(#)tcp_var.h 7.6 (Berkeley) 12/7/87 * * Modified for x-kernel v3.3 * Modifications Copyright (c) 1991,1996 Arizona Board of Regents * * $Revision: 1.3 $ * $Date: 1996/02/01 15:30:57 $ */ /* * xtcp configuration: This is a half-assed attempt to make xtcp * self-configure for a few varieties of 4.2 and 4.3-based unixes. * If you don't have a) a 4.3bsd vax or b) a 3.x Sun (x<6), check * this carefully (it's probably not right). Please send me mail * if you run into configuration problems. * - Van Jacobson (van@lbl-csam.arpa) */ #ifndef tcp_var_h #define tcp_var_h #include "tcp_timer.h" #ifndef BSD #define BSD 42 /* if we're not 4.3, pretend we're 4.2 */ #endif #if sun||BSD<43 #define TCP_COMPAT_42 /* set if we have to interop w/4.2 systems */ #endif #ifndef SB_MAX #ifdef SB_MAXCOUNT #define SB_MAX SB_MAXCOUNT /* Sun has to be a little bit different... */ #else #define SB_MAX 65535 /* XXX */ #endif SB_MAXCOUNT #endif SB_MAX /* * Bill Nowicki pointed out that the page size (CLBYTES) has * nothing to do with the mbuf cluster size. So, we followed * Sun's lead and made the new define MCLBYTES stand for the mbuf * cluster size. The following define makes up backwards compatible * with 4.3 and 4.2. If CLBYTES is >1024 on your machine, check * this against the mbuf cluster definitions in /usr/include/sys/mbuf.h. */ #ifndef MCLBYTES #define MCLBYTES CLBYTES /* XXX */ #endif /* * The routine in_localaddr is broken in Sun's 3.4. We redefine ours * (in tcp_input.c) so we use can it but won't have a name conflict. */ #ifdef sun #define in_localaddr tcp_in_localaddr #endif /* --------------- end of xtcp config ---------------- */ /* * Kernel variables for tcp. */ struct reass { struct reass *next, *prev; struct tcphdr th; Msg m; }; /* * Tcp control block, one per tcp; fields: */ struct tcpcb { struct reass *seg_next; /* sequencing queue */ struct reass *seg_prev; short t_state; /* state of this connection */ short t_timer[TCPT_NTIMERS]; /* tcp timers */ short t_rxtshift; /* log(2) of rexmt exp. backoff */ short t_rxtcur; /* current retransmit value */ short t_dupacks; /* consecutive dup acks recd */ int t_rexmtthresh; /* number of dupacks before retransmit */ int t_slowstart; /* number of packets to send after slowstart */ u_short t_maxseg; /* maximum segment size */ char t_force; /* 1 if forcing out a byte */ u_short t_flags; #define TF_ACKNOW 0x01 /* ack peer immediately */ #define TF_DELACK 0x02 /* ack, but try to delay it */ #define TF_NODELAY 0x04 /* don't delay packets to coalesce */ #define TF_NOOPT 0x08 /* don't use tcp options */ #define TF_SENTFIN 0x10 /* have sent FIN */ #define TF_REQ_SCALE 0x0020 /* have/will request window scaling */ #define TF_RCVD_SCALE 0x0040 /* other side has requested scaling */ #define TF_REQ_TSTMP 0x0080 /* have/will request timestamps */ #define TF_RCVD_TSTMP 0x0100 /* a timestamp was received in SYN */ #define TF_SACK_PERMIT 0x0200 /* other side said I could SACK */ struct tcpiphdr *t_template; /* skeletal packet for transmit */ struct inpcb *t_inpcb; /* back pointer to internet pcb */ /* * The following fields are used as in the protocol specification. * See RFC783, Dec. 1981, page 21. */ /* send sequence variables */ tcp_seq snd_una; /* send unacknowledged */ tcp_seq snd_nxt; /* send next */ tcp_seq snd_up; /* send urgent pointer */ tcp_seq snd_wl1; /* window update seg seq number */ tcp_seq snd_wl2; /* window update seg ack number */ tcp_seq iss; /* initial send sequence number */ u_int snd_wnd; /* send window */ /* receive sequence variables */ u_int rcv_wnd; /* receive window */ tcp_seq rcv_nxt; /* receive next */ tcp_seq rcv_up; /* receive urgent pointer */ tcp_seq irs; /* initial receive sequence number */ /* * Additional variables for this implementation. */ /* receive variables */ tcp_seq rcv_adv; /* advertised window */ /* retransmit variables */ tcp_seq snd_max; /* highest sequence number sent * used to recognize retransmits */ /* congestion control (for slow start, source quench, retransmit after loss) */ u_int snd_cwnd; /* congestion-controlled window */ u_int snd_ssthresh; /* snd_cwnd size threshhold for * for slow start exponential to * linear switch */ /* * transmit timing stuff. * srtt and rttvar are stored as fixed point; for convenience in smoothing, * srtt has 3 bits to the right of the binary point, rttvar has 2. * "Variance" is actually smoothed difference. */ short t_idle; /* inactivity time */ short t_rtt; /* round trip time */ tcp_seq t_rtseq; /* sequence number being timed */ short t_srtt; /* smoothed round-trip time */ short t_rttvar; /* variance in round-trip time */ u_short t_rttmin; /* minimum rtt allowed */ u_short max_rcvd; /* most peer has sent into window */ u_int max_sndwnd; /* largest window peer has offered */ /* out-of-band data */ char t_oobflags; /* have some */ char t_iobc; /* input character */ #define TCPOOB_HAVEDATA 0x01 #define TCPOOB_HADDATA 0x02 short t_softerror; /* possible error not yet reported */ /* LSB: from 4.4 */ /* RFC 1323 variables */ u_char snd_scale; /* window scaling for send window */ u_char rcv_scale; /* window scaling for recv window */ u_char request_r_scale; /* pending window scaling */ u_char requested_s_scale; u_int ts_recent; /* timestamp echo data */ u_int ts_recent_age; /* when last updated */ tcp_seq last_ack_sent; /* Added to test delayed acks vs. immediate acks */ int delack; int win_incr2; /* Vegas timer stuff */ int v_rtt; int v_sa; int v_sd; int v_timeout; int v_worried; /* While > 0, check for packet * timeout */ int v_time_cwin_chg; /* Time when congestion window * was last decreased. Used to prevent * decreasing cwin more than once * for losses which occurred at * same cwin level */ int v_synsent; /* Time when SYN was sent. Used to * set v_rtt at beg of connection */ int v_newcwnd; /* Stores new uninflated cwnd, used * to reset cwnd later. Same function * as snd_ssthresh, but now * snd_ssthresh is only used during * slow-start */ /* Vegas exp decrease in rate. Option to decrease window exponentially * (sends on every other ACK) rather than waiting 1/2 RTT to start * sending. NOT USED anymore (better to be conservative) */ int v_send_toggle; int v_use_sndwnd; int v_max_send; /* New Vegas stuff (12/14/93) */ /* Protect from spikes. Limits output rate to at most twice the * current rate. * It helps for fairness, and results in a little less losses, * but there is a problem when the receiver doesn't ACK every * packet immediately (delayed ACKs) resulting in the sender * not being able to keep up with the congestion window. * Therefore it is OFF by default, and my be removed in the final * version. */ int v_spike_do; int v_spike_nextTime; int v_spike_timeInc; int v_spike_byteCnt; /* Exp increase every other during slow-start. */ int v_exp_inc_do; int v_exp_inc_flag; int v_exp_inc_cnt; int v_exp_inc_isless; int v_exp_inc_nseg; /* Congestion detection (or more correctly avoidance ) variables. * The mechanism checks once per RTT. Current RTT began with sequence * v_cong_detect_begseq */ int v_cong_detect_do; /* Whether to do it */ tcp_seq v_cong_detect_begseq; int v_cong_detect_begtime; /* When RTT began */ int v_cong_detect_predict_do; /* Whether to predict the * available bw during * slow-start */ int v_cong_detect_last_sendtime; int v_cong_detect_baseRtt; /* Current base RTT, usually * the minimum */ int v_cong_detect_sumRTT; /* Used to get average */ int v_cong_detect_cntRTT; /* RTT during a RTT interval */ /* The congestion window should stay within band. Going above * it means that not enough buffers are being used at the * bottleneck, so cwnd should increase. Going below the band means * that too many buffers are being used, so cwnd should decrease. */ int v_cong_detect_top_isless; /* => cwnd < top of band */ int v_cong_detect_top_nseg; int v_cong_detect_bot_isless; /* => cwnd < bot of band */ int v_cong_detect_bot_nseg; int v_incr; /* How much to increase cwnd * when an ACK is received */ int v_other1; /* For testing new ideas, so don't need */ int v_other2; /* to recompile everything due to changing */ int v_other3; /* this .h file */ int v_other4; int v_other5; int v_other6; int v_other7; int v_other8; int v_other9; int v_other10; int v_other11; int v_other12; int v_other13; int v_other14; }; #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) /* LSB: from 4.4 */ /* * The smoothed round-trip time and estimated variance * are stored as fixed point numbers scaled by the values below. * For convenience, these scales are also used in smoothing the average * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). * With these scales, srtt has 3 bits to the right of the binary point, * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the * binary point, and is smoothed with an ALPHA of 0.75. */ #define TCP_RTT_SCALE 8 /* multiplier for srtt; 3 bits frac. */ #define TCP_RTT_SHIFT 3 /* shift for srtt; 3 bits frac. */ #define TCP_RTTVAR_SCALE 4 /* multiplier for rttvar; 2 bits */ #define TCP_RTTVAR_SHIFT 2 /* multiplier for rttvar; 2 bits */ /* * The initial retransmission should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar * will each average +1/2 tick of bias. When we compute * the retransmit timer, we want 1/2 tick of rounding and * 1 extra tick because of +-1/2 tick uncertainty in the * firing of the timer. The bias will give us exactly the * 1.5 tick we need. But, because the bias is * statistical, we have to test that we don't drop below * the minimum feasible timer (which is 2 ticks). * This macro assumes that the value of TCP_RTTVAR_SCALE * is the same as the multiplier for rttvar. */ #define TCP_REXMTVAL(tp) \ (((tp)->t_srtt >> TCP_RTT_SHIFT) + (tp)->t_rttvar) /* * TCP statistics. * Many of these should be kept per connection, * but that's inconvenient at the moment. */ struct tcpstat { #if BSD<=43 int tcps_badsum; int tcps_badoff; int tcps_hdrops; int tcps_badsegs; int tcps_unack; /* 4.3+ BSD stats start here */ #endif u_int tcps_connattempt; /* connections initiated */ u_int tcps_accepts; /* connections accepted */ u_int tcps_connects; /* connections established */ u_int tcps_drops; /* connections dropped */ u_int tcps_conndrops; /* embryonic connections dropped */ u_int tcps_closed; /* conn. closed (includes drops) */ u_int tcps_segstimed; /* segs where we tried to get rtt */ u_int tcps_rttupdated; /* times we succeeded */ u_int tcps_delack; /* delayed acks sent */ u_int tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ u_int tcps_rexmttimeo; /* retransmit timeouts */ u_int tcps_persisttimeo; /* persist timeouts */ u_int tcps_keeptimeo; /* keepalive timeouts */ u_int tcps_keepprobe; /* keepalive probes sent */ u_int tcps_keepdrops; /* connections dropped in keepalive */ u_int tcps_sndtotal; /* total packets sent */ u_int tcps_sndpack; /* data packets sent */ u_int tcps_sndbyte; /* data bytes sent */ u_int tcps_sndrexmitpack; /* data packets retransmitted */ u_int tcps_sndrexmitbyte; /* data bytes retransmitted */ u_int tcps_sndacks; /* ack-only packets sent */ u_int tcps_sndprobe; /* window probes sent */ u_int tcps_sndurg; /* packets sent with URG only */ u_int tcps_sndwinup; /* window update-only packets sent */ u_int tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ u_int tcps_rcvtotal; /* total packets received */ u_int tcps_rcvpack; /* packets received in sequence */ u_int tcps_rcvbyte; /* bytes received in sequence */ u_int tcps_rcvbadsum; /* packets received with ccksum errs */ u_int tcps_rcvbadoff; /* packets received with bad offset */ u_int tcps_rcvshort; /* packets received too short */ u_int tcps_rcvduppack; /* duplicate-only packets received */ u_int tcps_rcvdupbyte; /* duplicate-only bytes received */ u_int tcps_rcvpartduppack; /* packets with some duplicate data */ u_int tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ u_int tcps_rcvoopack; /* out-of-order packets received */ u_int tcps_rcvoobyte; /* out-of-order bytes received */ u_int tcps_rcvpackafterwin; /* packets with data after window */ u_int tcps_rcvbyteafterwin; /* bytes rcvd after window */ u_int tcps_rcvafterclose; /* packets rcvd after "close" */ u_int tcps_rcvwinprobe; /* rcvd window probe packets */ u_int tcps_rcvdupack; /* rcvd duplicate acks */ u_int tcps_rcvacktoomuch; /* rcvd acks for unsent data */ u_int tcps_rcvackpack; /* rcvd ack packets */ u_int tcps_rcvackbyte; /* bytes acked by rcvd acks */ u_int tcps_rcvwinupd; /* rcvd window update packets */ /* LSB: from 4.4 */ u_int tcps_pawsdrop; /* segments dropped due to PAWS */ u_int tcps_predack; /* times hdr predict ok for acks */ u_int tcps_preddat; /*times hdr predict ok for data pkts */ u_int tcps_pcbcachemiss; /* LSB: for Vegas */ u_int tcps_dupftimeo; /* Fine grained timeout on dup ACK */ u_int tcps_worryftimeo; /* Fine grained timeout on worry */ }; /*struct inpcb tcb;*/ /* head of queue of active tcpcb's */ /*struct tcpstat tcpstat;*/ /* tcp statistics */ struct tcpiphdr *tcp_template(struct tcpcb *); struct tcpcb *tcp_destroy(struct tcpcb *), *tcp_drop(struct tcpcb *, int); /* struct tcpcb *tcp_timers(); */ struct tcpcb *tcp_disconnect( struct tcpcb *); struct tcpcb *tcp_usrclosed( struct tcpcb *); #endif