tcp_var.h revision 330897
1/*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1993, 1994, 1995 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 4. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95 32 * $FreeBSD: stable/11/sys/netinet/tcp_var.h 330897 2018-03-14 03:19:51Z eadler $ 33 */ 34 35#ifndef _NETINET_TCP_VAR_H_ 36#define _NETINET_TCP_VAR_H_ 37 38#include <netinet/tcp.h> 39#include <netinet/tcp_fsm.h> 40 41#ifdef _KERNEL 42#include <net/vnet.h> 43#include <sys/mbuf.h> 44 45/* 46 * Kernel variables for tcp. 47 */ 48VNET_DECLARE(int, tcp_do_rfc1323); 49#define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323) 50 51#endif /* _KERNEL */ 52 53/* TCP segment queue entry */ 54struct tseg_qent { 55 LIST_ENTRY(tseg_qent) tqe_q; 56 int tqe_len; /* TCP segment data length */ 57 struct tcphdr *tqe_th; /* a pointer to tcp header */ 58 struct mbuf *tqe_m; /* mbuf contains packet */ 59}; 60LIST_HEAD(tsegqe_head, tseg_qent); 61 62struct sackblk { 63 tcp_seq start; /* start seq no. of sack block */ 64 tcp_seq end; /* end seq no. */ 65}; 66 67struct sackhole { 68 tcp_seq start; /* start seq no. of hole */ 69 tcp_seq end; /* end seq no. */ 70 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ 71 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ 72}; 73 74struct sackhint { 75 struct sackhole *nexthole; 76 int sack_bytes_rexmit; 77 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */ 78 79 int ispare; /* explicit pad for 64bit alignment */ 80 int sacked_bytes; /* 81 * Total sacked bytes reported by the 82 * receiver via sack option 83 */ 84 uint32_t _pad1[1]; /* TBD */ 85 uint64_t _pad[1]; /* TBD */ 86}; 87 88struct tcptemp { 89 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ 90 struct tcphdr tt_t; 91}; 92 93#define tcp6cb tcpcb /* for KAME src sync over BSD*'s */ 94 95/* 96 * TODO: We yet need to brave plowing in 97 * to tcp_input() and the pru_usrreq() block. 98 * Right now these go to the old standards which 99 * are somewhat ok, but in the long term may 100 * need to be changed. If we do tackle tcp_input() 101 * then we need to get rid of the tcp_do_segment() 102 * function below. 103 */ 104/* Flags for tcp functions */ 105#define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */ 106struct tcpcb; 107struct inpcb; 108struct sockopt; 109struct socket; 110 111/* 112 * If defining the optional tcp_timers, in the 113 * tfb_tcp_timer_stop call you must use the 114 * callout_async_drain() function with the 115 * tcp_timer_discard callback. You should check 116 * the return of callout_async_drain() and if 0 117 * increment tt_draincnt. Since the timer sub-system 118 * does not know your callbacks you must provide a 119 * stop_all function that loops through and calls 120 * tcp_timer_stop() with each of your defined timers. 121 */ 122struct tcp_function_block { 123 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX]; 124 int (*tfb_tcp_output)(struct tcpcb *); 125 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *, 126 struct socket *, struct tcpcb *, 127 int, int, uint8_t, 128 int); 129 int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt, 130 struct inpcb *inp, struct tcpcb *tp); 131 /* Optional memory allocation/free routine */ 132 void (*tfb_tcp_fb_init)(struct tcpcb *); 133 void (*tfb_tcp_fb_fini)(struct tcpcb *); 134 /* Optional timers, must define all if you define one */ 135 int (*tfb_tcp_timer_stop_all)(struct tcpcb *); 136 void (*tfb_tcp_timer_activate)(struct tcpcb *, 137 uint32_t, u_int); 138 int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t); 139 void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t); 140 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *); 141 volatile uint32_t tfb_refcnt; 142 uint32_t tfb_flags; 143}; 144 145struct tcp_function { 146 TAILQ_ENTRY(tcp_function) tf_next; 147 struct tcp_function_block *tf_fb; 148}; 149 150TAILQ_HEAD(tcp_funchead, tcp_function); 151 152/* 153 * Tcp control block, one per tcp; fields: 154 * Organized for 16 byte cacheline efficiency. 155 */ 156struct tcpcb { 157 struct tsegqe_head t_segq; /* segment reassembly queue */ 158 void *t_pspare[2]; /* new reassembly queue */ 159 int t_segqlen; /* segment reassembly queue length */ 160 int t_dupacks; /* consecutive dup acks recd */ 161 162 struct tcp_timer *t_timers; /* All the TCP timers in one struct */ 163 164 struct inpcb *t_inpcb; /* back pointer to internet pcb */ 165 int t_state; /* state of this connection */ 166 u_int t_flags; 167 168 struct vnet *t_vnet; /* back pointer to parent vnet */ 169 170 tcp_seq snd_una; /* sent but unacknowledged */ 171 tcp_seq snd_max; /* highest sequence number sent; 172 * used to recognize retransmits 173 */ 174 tcp_seq snd_nxt; /* send next */ 175 tcp_seq snd_up; /* send urgent pointer */ 176 177 tcp_seq snd_wl1; /* window update seg seq number */ 178 tcp_seq snd_wl2; /* window update seg ack number */ 179 tcp_seq iss; /* initial send sequence number */ 180 tcp_seq irs; /* initial receive sequence number */ 181 182 tcp_seq rcv_nxt; /* receive next */ 183 tcp_seq rcv_adv; /* advertised window */ 184 u_long rcv_wnd; /* receive window */ 185 tcp_seq rcv_up; /* receive urgent pointer */ 186 187 u_long snd_wnd; /* send window */ 188 u_long snd_cwnd; /* congestion-controlled window */ 189 u_long snd_spare1; /* unused */ 190 u_long snd_ssthresh; /* snd_cwnd size threshold for 191 * for slow start exponential to 192 * linear switch 193 */ 194 u_long snd_spare2; /* unused */ 195 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ 196 197 u_int t_rcvtime; /* inactivity time */ 198 u_int t_starttime; /* time connection was established */ 199 u_int t_rtttime; /* RTT measurement start time */ 200 tcp_seq t_rtseq; /* sequence number being timed */ 201 202 u_int t_bw_spare1; /* unused */ 203 tcp_seq t_bw_spare2; /* unused */ 204 205 int t_rxtcur; /* current retransmit value (ticks) */ 206 u_int t_maxseg; /* maximum segment size */ 207 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */ 208 int t_srtt; /* smoothed round-trip time */ 209 int t_rttvar; /* variance in round-trip time */ 210 211 int t_rxtshift; /* log(2) of rexmt exp. backoff */ 212 u_int t_rttmin; /* minimum rtt allowed */ 213 u_int t_rttbest; /* best rtt we've seen */ 214 u_long t_rttupdated; /* number of times rtt sampled */ 215 u_long max_sndwnd; /* largest window peer has offered */ 216 217 int t_softerror; /* possible error not yet reported */ 218/* out-of-band data */ 219 char t_oobflags; /* have some */ 220 char t_iobc; /* input character */ 221/* RFC 1323 variables */ 222 u_char snd_scale; /* window scaling for send window */ 223 u_char rcv_scale; /* window scaling for recv window */ 224 u_char request_r_scale; /* pending window scaling */ 225 u_int32_t ts_recent; /* timestamp echo data */ 226 u_int ts_recent_age; /* when last updated */ 227 u_int32_t ts_offset; /* our timestamp offset */ 228 229 tcp_seq last_ack_sent; 230/* experimental */ 231 u_long snd_cwnd_prev; /* cwnd prior to retransmit */ 232 u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */ 233 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ 234 int t_sndzerowin; /* zero-window updates sent */ 235 u_int t_badrxtwin; /* window for retransmit recovery */ 236 u_char snd_limited; /* segments limited transmitted */ 237/* SACK related state */ 238 int snd_numholes; /* number of holes seen by sender */ 239 TAILQ_HEAD(sackhole_head, sackhole) snd_holes; 240 /* SACK scoreboard (sorted) */ 241 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ 242 int rcv_numsacks; /* # distinct sack blks present */ 243 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ 244 tcp_seq sack_newdata; /* New data xmitted in this recovery 245 episode starts at this seq number */ 246 struct sackhint sackhint; /* SACK scoreboard hint */ 247 int t_rttlow; /* smallest observerved RTT */ 248 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ 249 int rfbuf_cnt; /* recv buffer autoscaling byte count */ 250 struct toedev *tod; /* toedev handling this connection */ 251 int t_sndrexmitpack; /* retransmit packets sent */ 252 int t_rcvoopack; /* out-of-order packets received */ 253 void *t_toe; /* TOE pcb pointer */ 254 int t_bytes_acked; /* # bytes acked during current RTT */ 255 struct cc_algo *cc_algo; /* congestion control algorithm */ 256 struct cc_var *ccv; /* congestion control specific vars */ 257 struct osd *osd; /* storage for Khelp module data */ 258 259 u_int t_keepinit; /* time to establish connection */ 260 u_int t_keepidle; /* time before keepalive probes begin */ 261 u_int t_keepintvl; /* interval between keepalives */ 262 u_int t_keepcnt; /* number of keepalives before close */ 263 264 u_int t_tsomax; /* TSO total burst length limit in bytes */ 265 u_int t_tsomaxsegcount; /* TSO maximum segment count */ 266 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */ 267 u_int t_flags2; /* More tcpcb flags storage */ 268#if defined(_KERNEL) && defined(TCP_RFC7413) 269 uint32_t t_ispare[6]; /* 5 UTO, 1 TBD */ 270 uint64_t t_tfo_cookie; /* TCP Fast Open cookie */ 271#else 272 uint32_t t_ispare[8]; /* 5 UTO, 3 TBD */ 273#endif 274 struct tcp_function_block *t_fb;/* TCP function call block */ 275 void *t_fb_ptr; /* Pointer to t_fb specific data */ 276#if defined(_KERNEL) && defined(TCP_RFC7413) 277 unsigned int *t_tfo_pending; /* TCP Fast Open pending counter */ 278 void *t_pspare2[1]; /* 1 TCP_SIGNATURE */ 279#else 280 void *t_pspare2[2]; /* 1 TCP_SIGNATURE, 1 TBD */ 281#endif 282#if defined(_KERNEL) && defined(TCPPCAP) 283 struct mbufq t_inpkts; /* List of saved input packets. */ 284 struct mbufq t_outpkts; /* List of saved output packets. */ 285#ifdef _LP64 286 uint64_t _pad[0]; /* all used! */ 287#else 288 uint64_t _pad[2]; /* 2 are available */ 289#endif /* _LP64 */ 290#else 291 uint64_t _pad[6]; 292#endif /* defined(_KERNEL) && defined(TCPPCAP) */ 293}; 294 295/* 296 * Flags and utility macros for the t_flags field. 297 */ 298#define TF_ACKNOW 0x000001 /* ack peer immediately */ 299#define TF_DELACK 0x000002 /* ack, but try to delay it */ 300#define TF_NODELAY 0x000004 /* don't delay packets to coalesce */ 301#define TF_NOOPT 0x000008 /* don't use tcp options */ 302#define TF_SENTFIN 0x000010 /* have sent FIN */ 303#define TF_REQ_SCALE 0x000020 /* have/will request window scaling */ 304#define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */ 305#define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */ 306#define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */ 307#define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */ 308#define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */ 309#define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */ 310#define TF_NOPUSH 0x001000 /* don't push */ 311#define TF_PREVVALID 0x002000 /* saved values for bad rxmit valid */ 312#define TF_MORETOCOME 0x010000 /* More data to be appended to sock */ 313#define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */ 314#define TF_LASTIDLE 0x040000 /* connection was previously idle */ 315#define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */ 316#define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */ 317#define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */ 318#define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */ 319#define TF_FORCEDATA 0x800000 /* force out a byte */ 320#define TF_TSO 0x1000000 /* TSO enabled on this connection */ 321#define TF_TOE 0x2000000 /* this connection is offloaded */ 322#define TF_ECN_PERMIT 0x4000000 /* connection ECN-ready */ 323#define TF_ECN_SND_CWR 0x8000000 /* ECN CWR in queue */ 324#define TF_ECN_SND_ECE 0x10000000 /* ECN ECE in queue */ 325#define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */ 326#define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */ 327#define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */ 328 329#define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY) 330#define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY 331#define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY 332 333#define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY) 334#define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY 335#define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY 336 337#define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY)) 338#define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY) 339#define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY) 340 341#define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una) 342 343/* 344 * Flags for the t_oobflags field. 345 */ 346#define TCPOOB_HAVEDATA 0x01 347#define TCPOOB_HADDATA 0x02 348 349/* 350 * Flags for PLPMTU handling, t_flags2 351 */ 352#define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */ 353#define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */ 354#define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */ 355 356/* 357 * Structure to hold TCP options that are only used during segment 358 * processing (in tcp_input), but not held in the tcpcb. 359 * It's basically used to reduce the number of parameters 360 * to tcp_dooptions and tcp_addoptions. 361 * The binary order of the to_flags is relevant for packing of the 362 * options in tcp_addoptions. 363 */ 364struct tcpopt { 365 u_int32_t to_flags; /* which options are present */ 366#define TOF_MSS 0x0001 /* maximum segment size */ 367#define TOF_SCALE 0x0002 /* window scaling */ 368#define TOF_SACKPERM 0x0004 /* SACK permitted */ 369#define TOF_TS 0x0010 /* timestamp */ 370#define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */ 371#define TOF_SACK 0x0080 /* Peer sent SACK option */ 372#define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */ 373#define TOF_MAXOPT 0x0200 374 u_int32_t to_tsval; /* new timestamp */ 375 u_int32_t to_tsecr; /* reflected timestamp */ 376 u_char *to_sacks; /* pointer to the first SACK blocks */ 377 u_char *to_signature; /* pointer to the TCP-MD5 signature */ 378 u_char *to_tfo_cookie; /* pointer to the TFO cookie */ 379 u_int16_t to_mss; /* maximum segment size */ 380 u_int8_t to_wscale; /* window scaling */ 381 u_int8_t to_nsacks; /* number of SACK blocks */ 382 u_int8_t to_tfo_len; /* TFO cookie length */ 383 u_int32_t to_spare; /* UTO */ 384}; 385 386/* 387 * Flags for tcp_dooptions. 388 */ 389#define TO_SYN 0x01 /* parse SYN-only options */ 390 391struct hc_metrics_lite { /* must stay in sync with hc_metrics */ 392 u_long rmx_mtu; /* MTU for this path */ 393 u_long rmx_ssthresh; /* outbound gateway buffer limit */ 394 u_long rmx_rtt; /* estimated round trip time */ 395 u_long rmx_rttvar; /* estimated rtt variance */ 396 u_long rmx_cwnd; /* congestion window */ 397 u_long rmx_sendpipe; /* outbound delay-bandwidth product */ 398 u_long rmx_recvpipe; /* inbound delay-bandwidth product */ 399}; 400 401/* 402 * Used by tcp_maxmtu() to communicate interface specific features 403 * and limits at the time of connection setup. 404 */ 405struct tcp_ifcap { 406 int ifcap; 407 u_int tsomax; 408 u_int tsomaxsegcount; 409 u_int tsomaxsegsize; 410}; 411 412#ifndef _NETINET_IN_PCB_H_ 413struct in_conninfo; 414#endif /* _NETINET_IN_PCB_H_ */ 415 416struct tcptw { 417 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */ 418 tcp_seq snd_nxt; 419 tcp_seq rcv_nxt; 420 tcp_seq iss; 421 tcp_seq irs; 422 u_short last_win; /* cached window value */ 423 u_short tw_so_options; /* copy of so_options */ 424 struct ucred *tw_cred; /* user credentials */ 425 u_int32_t t_recent; 426 u_int32_t ts_offset; /* our timestamp offset */ 427 u_int t_starttime; 428 int tw_time; 429 TAILQ_ENTRY(tcptw) tw_2msl; 430 void *tw_pspare; /* TCP_SIGNATURE */ 431 u_int *tw_spare; /* TCP_SIGNATURE */ 432}; 433 434#define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) 435#define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb) 436#define sototcpcb(so) (intotcpcb(sotoinpcb(so))) 437 438/* 439 * The smoothed round-trip time and estimated variance 440 * are stored as fixed point numbers scaled by the values below. 441 * For convenience, these scales are also used in smoothing the average 442 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). 443 * With these scales, srtt has 3 bits to the right of the binary point, 444 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the 445 * binary point, and is smoothed with an ALPHA of 0.75. 446 */ 447#define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ 448#define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ 449#define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ 450#define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ 451#define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ 452 453/* 454 * The initial retransmission should happen at rtt + 4 * rttvar. 455 * Because of the way we do the smoothing, srtt and rttvar 456 * will each average +1/2 tick of bias. When we compute 457 * the retransmit timer, we want 1/2 tick of rounding and 458 * 1 extra tick because of +-1/2 tick uncertainty in the 459 * firing of the timer. The bias will give us exactly the 460 * 1.5 tick we need. But, because the bias is 461 * statistical, we have to test that we don't drop below 462 * the minimum feasible timer (which is 2 ticks). 463 * This version of the macro adapted from a paper by Lawrence 464 * Brakmo and Larry Peterson which outlines a problem caused 465 * by insufficient precision in the original implementation, 466 * which results in inappropriately large RTO values for very 467 * fast networks. 468 */ 469#define TCP_REXMTVAL(tp) \ 470 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ 471 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) 472 473/* 474 * TCP statistics. 475 * Many of these should be kept per connection, 476 * but that's inconvenient at the moment. 477 */ 478struct tcpstat { 479 uint64_t tcps_connattempt; /* connections initiated */ 480 uint64_t tcps_accepts; /* connections accepted */ 481 uint64_t tcps_connects; /* connections established */ 482 uint64_t tcps_drops; /* connections dropped */ 483 uint64_t tcps_conndrops; /* embryonic connections dropped */ 484 uint64_t tcps_minmssdrops; /* average minmss too low drops */ 485 uint64_t tcps_closed; /* conn. closed (includes drops) */ 486 uint64_t tcps_segstimed; /* segs where we tried to get rtt */ 487 uint64_t tcps_rttupdated; /* times we succeeded */ 488 uint64_t tcps_delack; /* delayed acks sent */ 489 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ 490 uint64_t tcps_rexmttimeo; /* retransmit timeouts */ 491 uint64_t tcps_persisttimeo; /* persist timeouts */ 492 uint64_t tcps_keeptimeo; /* keepalive timeouts */ 493 uint64_t tcps_keepprobe; /* keepalive probes sent */ 494 uint64_t tcps_keepdrops; /* connections dropped in keepalive */ 495 496 uint64_t tcps_sndtotal; /* total packets sent */ 497 uint64_t tcps_sndpack; /* data packets sent */ 498 uint64_t tcps_sndbyte; /* data bytes sent */ 499 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */ 500 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */ 501 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */ 502 uint64_t tcps_sndacks; /* ack-only packets sent */ 503 uint64_t tcps_sndprobe; /* window probes sent */ 504 uint64_t tcps_sndurg; /* packets sent with URG only */ 505 uint64_t tcps_sndwinup; /* window update-only packets sent */ 506 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ 507 508 uint64_t tcps_rcvtotal; /* total packets received */ 509 uint64_t tcps_rcvpack; /* packets received in sequence */ 510 uint64_t tcps_rcvbyte; /* bytes received in sequence */ 511 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */ 512 uint64_t tcps_rcvbadoff; /* packets received with bad offset */ 513 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */ 514 uint64_t tcps_rcvshort; /* packets received too short */ 515 uint64_t tcps_rcvduppack; /* duplicate-only packets received */ 516 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */ 517 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */ 518 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ 519 uint64_t tcps_rcvoopack; /* out-of-order packets received */ 520 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */ 521 uint64_t tcps_rcvpackafterwin; /* packets with data after window */ 522 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */ 523 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */ 524 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */ 525 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */ 526 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */ 527 uint64_t tcps_rcvackpack; /* rcvd ack packets */ 528 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */ 529 uint64_t tcps_rcvwinupd; /* rcvd window update packets */ 530 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */ 531 uint64_t tcps_predack; /* times hdr predict ok for acks */ 532 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */ 533 uint64_t tcps_pcbcachemiss; 534 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */ 535 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */ 536 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */ 537 uint64_t tcps_usedrtt; /* times RTT initialized from route */ 538 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */ 539 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/ 540 uint64_t tcps_persistdrop; /* timeout in persist state */ 541 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */ 542 uint64_t tcps_mturesent; /* resends due to MTU discovery */ 543 uint64_t tcps_listendrop; /* listen queue overflows */ 544 uint64_t tcps_badrst; /* ignored RSTs in the window */ 545 546 uint64_t tcps_sc_added; /* entry added to syncache */ 547 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */ 548 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */ 549 uint64_t tcps_sc_dropped; /* could not reply to packet */ 550 uint64_t tcps_sc_completed; /* successful extraction of entry */ 551 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */ 552 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */ 553 uint64_t tcps_sc_reset; /* RST removed entry from syncache */ 554 uint64_t tcps_sc_stale; /* timed out or listen socket gone */ 555 uint64_t tcps_sc_aborted; /* syncache entry aborted */ 556 uint64_t tcps_sc_badack; /* removed due to bad ACK */ 557 uint64_t tcps_sc_unreach; /* ICMP unreachable received */ 558 uint64_t tcps_sc_zonefail; /* zalloc() failed */ 559 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */ 560 uint64_t tcps_sc_recvcookie; /* SYN cookie received */ 561 562 uint64_t tcps_hc_added; /* entry added to hostcache */ 563 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */ 564 565 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ 566 567 /* SACK related stats */ 568 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */ 569 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */ 570 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ 571 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ 572 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ 573 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */ 574 575 /* ECN related stats */ 576 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */ 577 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */ 578 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */ 579 uint64_t tcps_ecn_shs; /* ECN successful handshakes */ 580 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */ 581 582 /* TCP_SIGNATURE related stats */ 583 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */ 584 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */ 585 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */ 586 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */ 587 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */ 588 589 uint64_t _pad[12]; /* 6 UTO, 6 TBD */ 590}; 591 592#define tcps_rcvmemdrop tcps_rcvreassfull /* compat */ 593 594#ifdef _KERNEL 595#define TI_UNLOCKED 1 596#define TI_RLOCKED 2 597#include <sys/counter.h> 598 599VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */ 600/* 601 * In-kernel consumers can use these accessor macros directly to update 602 * stats. 603 */ 604#define TCPSTAT_ADD(name, val) \ 605 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)) 606#define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1) 607 608/* 609 * Kernel module consumers must use this accessor macro. 610 */ 611void kmod_tcpstat_inc(int statnum); 612#define KMOD_TCPSTAT_INC(name) \ 613 kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t)) 614 615/* 616 * Running TCP connection count by state. 617 */ 618VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]); 619#define V_tcps_states VNET(tcps_states) 620#define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1) 621#define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1) 622 623/* 624 * TCP specific helper hook point identifiers. 625 */ 626#define HHOOK_TCP_EST_IN 0 627#define HHOOK_TCP_EST_OUT 1 628#define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT 629 630struct tcp_hhook_data { 631 struct tcpcb *tp; 632 struct tcphdr *th; 633 struct tcpopt *to; 634 long len; 635 int tso; 636 tcp_seq curack; 637}; 638#endif 639 640/* 641 * TCB structure exported to user-land via sysctl(3). 642 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been 643 * included. Not all of our clients do. 644 */ 645#if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) 646struct xtcp_timer { 647 int tt_rexmt; /* retransmit timer */ 648 int tt_persist; /* retransmit persistence */ 649 int tt_keep; /* keepalive */ 650 int tt_2msl; /* 2*msl TIME_WAIT timer */ 651 int tt_delack; /* delayed ACK timer */ 652 int t_rcvtime; /* Time since last packet received */ 653}; 654struct xtcpcb { 655 size_t xt_len; 656 struct inpcb xt_inp; 657 struct tcpcb xt_tp; 658 struct xsocket xt_socket; 659 struct xtcp_timer xt_timer; 660 u_quad_t xt_alignment_hack; 661}; 662#endif 663 664/* 665 * Identifiers for TCP sysctl nodes 666 */ 667#define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ 668#define TCPCTL_MSSDFLT 3 /* MSS default */ 669#define TCPCTL_STATS 4 /* statistics */ 670#define TCPCTL_RTTDFLT 5 /* default RTT estimate */ 671#define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ 672#define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ 673#define TCPCTL_SENDSPACE 8 /* send buffer space */ 674#define TCPCTL_RECVSPACE 9 /* receive buffer space */ 675#define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ 676#define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ 677#define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ 678#define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ 679#define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ 680#define TCPCTL_DROP 15 /* drop tcp connection */ 681#define TCPCTL_STATES 16 /* connection counts by TCP state */ 682 683#ifdef _KERNEL 684#ifdef SYSCTL_DECL 685SYSCTL_DECL(_net_inet_tcp); 686SYSCTL_DECL(_net_inet_tcp_sack); 687MALLOC_DECLARE(M_TCPLOG); 688#endif 689 690VNET_DECLARE(struct inpcbhead, tcb); /* queue of active tcpcb's */ 691VNET_DECLARE(struct inpcbinfo, tcbinfo); 692extern int tcp_log_in_vain; 693VNET_DECLARE(int, tcp_mssdflt); /* XXX */ 694VNET_DECLARE(int, tcp_minmss); 695VNET_DECLARE(int, tcp_delack_enabled); 696VNET_DECLARE(int, tcp_do_rfc3390); 697VNET_DECLARE(int, tcp_initcwnd_segments); 698VNET_DECLARE(int, tcp_sendspace); 699VNET_DECLARE(int, tcp_recvspace); 700VNET_DECLARE(int, path_mtu_discovery); 701VNET_DECLARE(int, tcp_do_rfc3465); 702VNET_DECLARE(int, tcp_abc_l_var); 703#define V_tcb VNET(tcb) 704#define V_tcbinfo VNET(tcbinfo) 705#define V_tcp_mssdflt VNET(tcp_mssdflt) 706#define V_tcp_minmss VNET(tcp_minmss) 707#define V_tcp_delack_enabled VNET(tcp_delack_enabled) 708#define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390) 709#define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments) 710#define V_tcp_sendspace VNET(tcp_sendspace) 711#define V_tcp_recvspace VNET(tcp_recvspace) 712#define V_path_mtu_discovery VNET(path_mtu_discovery) 713#define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465) 714#define V_tcp_abc_l_var VNET(tcp_abc_l_var) 715 716VNET_DECLARE(int, tcp_do_sack); /* SACK enabled/disabled */ 717VNET_DECLARE(int, tcp_sc_rst_sock_fail); /* RST on sock alloc failure */ 718#define V_tcp_do_sack VNET(tcp_do_sack) 719#define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail) 720 721VNET_DECLARE(int, tcp_do_ecn); /* TCP ECN enabled/disabled */ 722VNET_DECLARE(int, tcp_ecn_maxretries); 723#define V_tcp_do_ecn VNET(tcp_do_ecn) 724#define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries) 725 726VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]); 727#define V_tcp_hhh VNET(tcp_hhh) 728 729VNET_DECLARE(int, tcp_do_rfc6675_pipe); 730#define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe) 731 732int tcp_addoptions(struct tcpopt *, u_char *); 733int tcp_ccalgounload(struct cc_algo *unload_algo); 734struct tcpcb * 735 tcp_close(struct tcpcb *); 736void tcp_discardcb(struct tcpcb *); 737void tcp_twstart(struct tcpcb *); 738void tcp_twclose(struct tcptw *, int); 739void tcp_ctlinput(int, struct sockaddr *, void *); 740int tcp_ctloutput(struct socket *, struct sockopt *); 741struct tcpcb * 742 tcp_drop(struct tcpcb *, int); 743void tcp_drain(void); 744void tcp_init(void); 745void tcp_fini(void *); 746char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *, 747 const void *); 748char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *, 749 const void *); 750int tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *); 751void tcp_reass_global_init(void); 752void tcp_reass_flush(struct tcpcb *); 753void tcp_dooptions(struct tcpopt *, u_char *, int, int); 754void tcp_dropwithreset(struct mbuf *, struct tcphdr *, 755 struct tcpcb *, int, int); 756void tcp_pulloutofband(struct socket *, 757 struct tcphdr *, struct mbuf *, int); 758void tcp_xmit_timer(struct tcpcb *, int); 759void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); 760void cc_ack_received(struct tcpcb *tp, struct tcphdr *th, 761 uint16_t type); 762void cc_conn_init(struct tcpcb *tp); 763void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); 764void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type); 765void hhook_run_tcp_est_in(struct tcpcb *tp, 766 struct tcphdr *th, struct tcpopt *to); 767 768int tcp_input(struct mbuf **, int *, int); 769int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *, 770 struct tcpcb *, int); 771void tcp_do_segment(struct mbuf *, struct tcphdr *, 772 struct socket *, struct tcpcb *, int, int, uint8_t, 773 int); 774 775int register_tcp_functions(struct tcp_function_block *blk, int wait); 776int deregister_tcp_functions(struct tcp_function_block *blk); 777struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs); 778struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk); 779int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp); 780 781u_long tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *); 782u_long tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *); 783u_int tcp_maxseg(const struct tcpcb *); 784void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *, 785 struct tcp_ifcap *); 786void tcp_mss(struct tcpcb *, int); 787int tcp_mssopt(struct in_conninfo *); 788struct inpcb * 789 tcp_drop_syn_sent(struct inpcb *, int); 790struct tcpcb * 791 tcp_newtcpcb(struct inpcb *); 792int tcp_output(struct tcpcb *); 793void tcp_state_change(struct tcpcb *, int); 794void tcp_respond(struct tcpcb *, void *, 795 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int); 796void tcp_tw_init(void); 797#ifdef VIMAGE 798void tcp_tw_destroy(void); 799#endif 800void tcp_tw_zone_change(void); 801int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *, 802 struct mbuf *, int); 803void tcp_setpersist(struct tcpcb *); 804void tcp_slowtimo(void); 805struct tcptemp * 806 tcpip_maketemplate(struct inpcb *); 807void tcpip_fillheaders(struct inpcb *, void *, void *); 808void tcp_timer_activate(struct tcpcb *, uint32_t, u_int); 809int tcp_timer_active(struct tcpcb *, uint32_t); 810void tcp_timer_stop(struct tcpcb *, uint32_t); 811void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int); 812/* 813 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) 814 */ 815void tcp_hc_init(void); 816#ifdef VIMAGE 817void tcp_hc_destroy(void); 818#endif 819void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); 820u_long tcp_hc_getmtu(struct in_conninfo *); 821void tcp_hc_updatemtu(struct in_conninfo *, u_long); 822void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); 823 824extern struct pr_usrreqs tcp_usrreqs; 825tcp_seq tcp_new_isn(struct tcpcb *); 826 827int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); 828void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); 829void tcp_clean_sackreport(struct tcpcb *tp); 830void tcp_sack_adjust(struct tcpcb *tp); 831struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); 832void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); 833void tcp_free_sackholes(struct tcpcb *tp); 834int tcp_newreno(struct tcpcb *, struct tcphdr *); 835int tcp_compute_pipe(struct tcpcb *); 836 837static inline void 838tcp_fields_to_host(struct tcphdr *th) 839{ 840 841 th->th_seq = ntohl(th->th_seq); 842 th->th_ack = ntohl(th->th_ack); 843 th->th_win = ntohs(th->th_win); 844 th->th_urp = ntohs(th->th_urp); 845} 846 847static inline void 848tcp_fields_to_net(struct tcphdr *th) 849{ 850 851 th->th_seq = htonl(th->th_seq); 852 th->th_ack = htonl(th->th_ack); 853 th->th_win = htons(th->th_win); 854 th->th_urp = htons(th->th_urp); 855} 856#endif /* _KERNEL */ 857 858#endif /* _NETINET_TCP_VAR_H_ */ 859