1 /*- 2 * Copyright (c) 1982, 1986, 1993, 1994, 1995 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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95 30 * $FreeBSD$ 31 */ 32 33 #ifndef _NETINET_TCP_VAR_H_ 34 #define _NETINET_TCP_VAR_H_ 35 36 #include <netinet/tcp.h> 37 #include <netinet/tcp_fsm.h> 38 39 #ifdef _KERNEL 40 #include <net/vnet.h> 41 #include <sys/mbuf.h> 42 43 /* 44 * Kernel variables for tcp. 45 */ 46 VNET_DECLARE(int, tcp_do_rfc1323); 47 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323) 48 49 #endif /* _KERNEL */ 50 51 /* TCP segment queue entry */ 52 struct tseg_qent { 53 LIST_ENTRY(tseg_qent) tqe_q; 54 int tqe_len; /* TCP segment data length */ 55 struct tcphdr *tqe_th; /* a pointer to tcp header */ 56 struct mbuf *tqe_m; /* mbuf contains packet */ 57 }; 58 LIST_HEAD(tsegqe_head, tseg_qent); 59 60 struct sackblk { 61 tcp_seq start; /* start seq no. of sack block */ 62 tcp_seq end; /* end seq no. */ 63 }; 64 65 struct sackhole { 66 tcp_seq start; /* start seq no. of hole */ 67 tcp_seq end; /* end seq no. */ 68 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ 69 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ 70 }; 71 72 struct sackhint { 73 struct sackhole *nexthole; 74 int sack_bytes_rexmit; 75 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */ 76 77 int ispare; /* explicit pad for 64bit alignment */ 78 int sacked_bytes; /* 79 * Total sacked bytes reported by the 80 * receiver via sack option 81 */ 82 uint32_t _pad1[1]; /* TBD */ 83 uint64_t _pad[1]; /* TBD */ 84 }; 85 86 struct tcptemp { 87 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ 88 struct tcphdr tt_t; 89 }; 90 91 #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */ 92 93 /* 94 * TODO: We yet need to brave plowing in 95 * to tcp_input() and the pru_usrreq() block. 96 * Right now these go to the old standards which 97 * are somewhat ok, but in the long term may 98 * need to be changed. If we do tackle tcp_input() 99 * then we need to get rid of the tcp_do_segment() 100 * function below. 101 */ 102 /* Flags for tcp functions */ 103 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */ 104 struct tcpcb; 105 struct inpcb; 106 struct sockopt; 107 struct socket; 108 109 /* 110 * If defining the optional tcp_timers, in the 111 * tfb_tcp_timer_stop call you must use the 112 * callout_async_drain() function with the 113 * tcp_timer_discard callback. You should check 114 * the return of callout_async_drain() and if 0 115 * increment tt_draincnt. Since the timer sub-system 116 * does not know your callbacks you must provide a 117 * stop_all function that loops through and calls 118 * tcp_timer_stop() with each of your defined timers. 119 * Adding a tfb_tcp_handoff_ok function allows the socket 120 * option to change stacks to query you even if the 121 * connection is in a later stage. You return 0 to 122 * say you can take over and run your stack, you return 123 * non-zero (an error number) to say no you can't. 124 * If the function is undefined you can only change 125 * in the early states (before connect or listen). 126 * tfb_tcp_fb_fini is changed to add a flag to tell 127 * the old stack if the tcb is being destroyed or 128 * not. A one in the flag means the TCB is being 129 * destroyed, a zero indicates its transitioning to 130 * another stack (via socket option). 131 */ 132 struct tcp_function_block { 133 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX]; 134 int (*tfb_tcp_output)(struct tcpcb *); 135 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *, 136 struct socket *, struct tcpcb *, 137 int, int, uint8_t, 138 int); 139 int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt, 140 struct inpcb *inp, struct tcpcb *tp); 141 /* Optional memory allocation/free routine */ 142 void (*tfb_tcp_fb_init)(struct tcpcb *); 143 void (*tfb_tcp_fb_fini)(struct tcpcb *, int); 144 /* Optional timers, must define all if you define one */ 145 int (*tfb_tcp_timer_stop_all)(struct tcpcb *); 146 void (*tfb_tcp_timer_activate)(struct tcpcb *, 147 uint32_t, u_int); 148 int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t); 149 void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t); 150 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *); 151 int (*tfb_tcp_handoff_ok)(struct tcpcb *); 152 volatile uint32_t tfb_refcnt; 153 uint32_t tfb_flags; 154 }; 155 156 struct tcp_function { 157 TAILQ_ENTRY(tcp_function) tf_next; 158 struct tcp_function_block *tf_fb; 159 }; 160 161 TAILQ_HEAD(tcp_funchead, tcp_function); 162 163 /* 164 * Tcp control block, one per tcp; fields: 165 * Organized for 16 byte cacheline efficiency. 166 */ 167 struct tcpcb { 168 struct tsegqe_head t_segq; /* segment reassembly queue */ 169 void *t_pspare[2]; /* new reassembly queue */ 170 int t_segqlen; /* segment reassembly queue length */ 171 int t_dupacks; /* consecutive dup acks recd */ 172 173 struct tcp_timer *t_timers; /* All the TCP timers in one struct */ 174 175 struct inpcb *t_inpcb; /* back pointer to internet pcb */ 176 int t_state; /* state of this connection */ 177 u_int t_flags; 178 179 struct vnet *t_vnet; /* back pointer to parent vnet */ 180 181 tcp_seq snd_una; /* sent but unacknowledged */ 182 tcp_seq snd_max; /* highest sequence number sent; 183 * used to recognize retransmits 184 */ 185 tcp_seq snd_nxt; /* send next */ 186 tcp_seq snd_up; /* send urgent pointer */ 187 188 tcp_seq snd_wl1; /* window update seg seq number */ 189 tcp_seq snd_wl2; /* window update seg ack number */ 190 tcp_seq iss; /* initial send sequence number */ 191 tcp_seq irs; /* initial receive sequence number */ 192 193 tcp_seq rcv_nxt; /* receive next */ 194 tcp_seq rcv_adv; /* advertised window */ 195 u_long rcv_wnd; /* receive window */ 196 tcp_seq rcv_up; /* receive urgent pointer */ 197 198 u_long snd_wnd; /* send window */ 199 u_long snd_cwnd; /* congestion-controlled window */ 200 u_long snd_spare1; /* unused */ 201 u_long snd_ssthresh; /* snd_cwnd size threshold for 202 * for slow start exponential to 203 * linear switch 204 */ 205 u_long snd_spare2; /* unused */ 206 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ 207 208 u_int t_rcvtime; /* inactivity time */ 209 u_int t_starttime; /* time connection was established */ 210 u_int t_rtttime; /* RTT measurement start time */ 211 tcp_seq t_rtseq; /* sequence number being timed */ 212 213 u_int t_bw_spare1; /* unused */ 214 tcp_seq t_bw_spare2; /* unused */ 215 216 int t_rxtcur; /* current retransmit value (ticks) */ 217 u_int t_maxseg; /* maximum segment size */ 218 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */ 219 int t_srtt; /* smoothed round-trip time */ 220 int t_rttvar; /* variance in round-trip time */ 221 222 int t_rxtshift; /* log(2) of rexmt exp. backoff */ 223 u_int t_rttmin; /* minimum rtt allowed */ 224 u_int t_rttbest; /* best rtt we've seen */ 225 u_long t_rttupdated; /* number of times rtt sampled */ 226 u_long max_sndwnd; /* largest window peer has offered */ 227 228 int t_softerror; /* possible error not yet reported */ 229 /* out-of-band data */ 230 char t_oobflags; /* have some */ 231 char t_iobc; /* input character */ 232 /* RFC 1323 variables */ 233 u_char snd_scale; /* window scaling for send window */ 234 u_char rcv_scale; /* window scaling for recv window */ 235 u_char request_r_scale; /* pending window scaling */ 236 u_int32_t ts_recent; /* timestamp echo data */ 237 u_int ts_recent_age; /* when last updated */ 238 u_int32_t ts_offset; /* our timestamp offset */ 239 240 tcp_seq last_ack_sent; 241 /* experimental */ 242 u_long snd_cwnd_prev; /* cwnd prior to retransmit */ 243 u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */ 244 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ 245 int t_sndzerowin; /* zero-window updates sent */ 246 u_int t_badrxtwin; /* window for retransmit recovery */ 247 u_char snd_limited; /* segments limited transmitted */ 248 /* SACK related state */ 249 int snd_numholes; /* number of holes seen by sender */ 250 TAILQ_HEAD(sackhole_head, sackhole) snd_holes; 251 /* SACK scoreboard (sorted) */ 252 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ 253 int rcv_numsacks; /* # distinct sack blks present */ 254 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ 255 tcp_seq sack_newdata; /* New data xmitted in this recovery 256 episode starts at this seq number */ 257 struct sackhint sackhint; /* SACK scoreboard hint */ 258 int t_rttlow; /* smallest observerved RTT */ 259 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ 260 int rfbuf_cnt; /* recv buffer autoscaling byte count */ 261 struct toedev *tod; /* toedev handling this connection */ 262 int t_sndrexmitpack; /* retransmit packets sent */ 263 int t_rcvoopack; /* out-of-order packets received */ 264 void *t_toe; /* TOE pcb pointer */ 265 int t_bytes_acked; /* # bytes acked during current RTT */ 266 struct cc_algo *cc_algo; /* congestion control algorithm */ 267 struct cc_var *ccv; /* congestion control specific vars */ 268 struct osd *osd; /* storage for Khelp module data */ 269 270 u_int t_keepinit; /* time to establish connection */ 271 u_int t_keepidle; /* time before keepalive probes begin */ 272 u_int t_keepintvl; /* interval between keepalives */ 273 u_int t_keepcnt; /* number of keepalives before close */ 274 275 u_int t_tsomax; /* TSO total burst length limit in bytes */ 276 u_int t_tsomaxsegcount; /* TSO maximum segment count */ 277 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */ 278 u_int t_flags2; /* More tcpcb flags storage */ 279 #if defined(_KERNEL) && defined(TCP_RFC7413) 280 uint32_t t_ispare[6]; /* 5 UTO, 1 TBD */ 281 uint64_t t_tfo_cookie; /* TCP Fast Open cookie */ 282 #else 283 uint32_t t_ispare[8]; /* 5 UTO, 3 TBD */ 284 #endif 285 struct tcp_function_block *t_fb;/* TCP function call block */ 286 void *t_fb_ptr; /* Pointer to t_fb specific data */ 287 #if defined(_KERNEL) && defined(TCP_RFC7413) 288 unsigned int *t_tfo_pending; /* TCP Fast Open pending counter */ 289 void *t_pspare2[1]; /* 1 TCP_SIGNATURE */ 290 #else 291 void *t_pspare2[2]; /* 1 TCP_SIGNATURE, 1 TBD */ 292 #endif 293 #if defined(_KERNEL) && defined(TCPPCAP) 294 struct mbufq t_inpkts; /* List of saved input packets. */ 295 struct mbufq t_outpkts; /* List of saved output packets. */ 296 #ifdef _LP64 297 uint64_t _pad[0]; /* all used! */ 298 #else 299 uint64_t _pad[2]; /* 2 are available */ 300 #endif /* _LP64 */ 301 #else 302 uint64_t _pad[6]; 303 #endif /* defined(_KERNEL) && defined(TCPPCAP) */ 304 }; 305 306 /* 307 * Flags and utility macros for the t_flags field. 308 */ 309 #define TF_ACKNOW 0x000001 /* ack peer immediately */ 310 #define TF_DELACK 0x000002 /* ack, but try to delay it */ 311 #define TF_NODELAY 0x000004 /* don't delay packets to coalesce */ 312 #define TF_NOOPT 0x000008 /* don't use tcp options */ 313 #define TF_SENTFIN 0x000010 /* have sent FIN */ 314 #define TF_REQ_SCALE 0x000020 /* have/will request window scaling */ 315 #define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */ 316 #define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */ 317 #define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */ 318 #define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */ 319 #define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */ 320 #define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */ 321 #define TF_NOPUSH 0x001000 /* don't push */ 322 #define TF_PREVVALID 0x002000 /* saved values for bad rxmit valid */ 323 #define TF_MORETOCOME 0x010000 /* More data to be appended to sock */ 324 #define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */ 325 #define TF_LASTIDLE 0x040000 /* connection was previously idle */ 326 #define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */ 327 #define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */ 328 #define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */ 329 #define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */ 330 #define TF_FORCEDATA 0x800000 /* force out a byte */ 331 #define TF_TSO 0x1000000 /* TSO enabled on this connection */ 332 #define TF_TOE 0x2000000 /* this connection is offloaded */ 333 #define TF_ECN_PERMIT 0x4000000 /* connection ECN-ready */ 334 #define TF_ECN_SND_CWR 0x8000000 /* ECN CWR in queue */ 335 #define TF_ECN_SND_ECE 0x10000000 /* ECN ECE in queue */ 336 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */ 337 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */ 338 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */ 339 340 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY) 341 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY 342 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY 343 344 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY) 345 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY 346 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY 347 348 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY)) 349 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY) 350 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY) 351 352 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una) 353 354 /* 355 * Flags for the t_oobflags field. 356 */ 357 #define TCPOOB_HAVEDATA 0x01 358 #define TCPOOB_HADDATA 0x02 359 360 #ifdef TCP_SIGNATURE 361 /* 362 * Defines which are needed by the xform_tcp module and tcp_[in|out]put 363 * for SADB verification and lookup. 364 */ 365 #define TCP_SIGLEN 16 /* length of computed digest in bytes */ 366 #define TCP_KEYLEN_MIN 1 /* minimum length of TCP-MD5 key */ 367 #define TCP_KEYLEN_MAX 80 /* maximum length of TCP-MD5 key */ 368 /* 369 * Only a single SA per host may be specified at this time. An SPI is 370 * needed in order for the KEY_ALLOCSA() lookup to work. 371 */ 372 #define TCP_SIG_SPI 0x1000 373 #endif /* TCP_SIGNATURE */ 374 375 /* 376 * Flags for PLPMTU handling, t_flags2 377 */ 378 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */ 379 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */ 380 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */ 381 382 /* 383 * Structure to hold TCP options that are only used during segment 384 * processing (in tcp_input), but not held in the tcpcb. 385 * It's basically used to reduce the number of parameters 386 * to tcp_dooptions and tcp_addoptions. 387 * The binary order of the to_flags is relevant for packing of the 388 * options in tcp_addoptions. 389 */ 390 struct tcpopt { 391 u_int32_t to_flags; /* which options are present */ 392 #define TOF_MSS 0x0001 /* maximum segment size */ 393 #define TOF_SCALE 0x0002 /* window scaling */ 394 #define TOF_SACKPERM 0x0004 /* SACK permitted */ 395 #define TOF_TS 0x0010 /* timestamp */ 396 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */ 397 #define TOF_SACK 0x0080 /* Peer sent SACK option */ 398 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */ 399 #define TOF_MAXOPT 0x0200 400 u_int32_t to_tsval; /* new timestamp */ 401 u_int32_t to_tsecr; /* reflected timestamp */ 402 u_char *to_sacks; /* pointer to the first SACK blocks */ 403 u_char *to_signature; /* pointer to the TCP-MD5 signature */ 404 u_char *to_tfo_cookie; /* pointer to the TFO cookie */ 405 u_int16_t to_mss; /* maximum segment size */ 406 u_int8_t to_wscale; /* window scaling */ 407 u_int8_t to_nsacks; /* number of SACK blocks */ 408 u_int8_t to_tfo_len; /* TFO cookie length */ 409 u_int32_t to_spare; /* UTO */ 410 }; 411 412 /* 413 * Flags for tcp_dooptions. 414 */ 415 #define TO_SYN 0x01 /* parse SYN-only options */ 416 417 struct hc_metrics_lite { /* must stay in sync with hc_metrics */ 418 u_long rmx_mtu; /* MTU for this path */ 419 u_long rmx_ssthresh; /* outbound gateway buffer limit */ 420 u_long rmx_rtt; /* estimated round trip time */ 421 u_long rmx_rttvar; /* estimated rtt variance */ 422 u_long rmx_cwnd; /* congestion window */ 423 u_long rmx_sendpipe; /* outbound delay-bandwidth product */ 424 u_long rmx_recvpipe; /* inbound delay-bandwidth product */ 425 }; 426 427 /* 428 * Used by tcp_maxmtu() to communicate interface specific features 429 * and limits at the time of connection setup. 430 */ 431 struct tcp_ifcap { 432 int ifcap; 433 u_int tsomax; 434 u_int tsomaxsegcount; 435 u_int tsomaxsegsize; 436 }; 437 438 #ifndef _NETINET_IN_PCB_H_ 439 struct in_conninfo; 440 #endif /* _NETINET_IN_PCB_H_ */ 441 442 struct tcptw { 443 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */ 444 tcp_seq snd_nxt; 445 tcp_seq rcv_nxt; 446 tcp_seq iss; 447 tcp_seq irs; 448 u_short last_win; /* cached window value */ 449 u_short tw_so_options; /* copy of so_options */ 450 struct ucred *tw_cred; /* user credentials */ 451 u_int32_t t_recent; 452 u_int32_t ts_offset; /* our timestamp offset */ 453 u_int t_starttime; 454 int tw_time; 455 TAILQ_ENTRY(tcptw) tw_2msl; 456 void *tw_pspare; /* TCP_SIGNATURE */ 457 u_int *tw_spare; /* TCP_SIGNATURE */ 458 }; 459 460 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) 461 #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb) 462 #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) 463 464 /* 465 * The smoothed round-trip time and estimated variance 466 * are stored as fixed point numbers scaled by the values below. 467 * For convenience, these scales are also used in smoothing the average 468 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). 469 * With these scales, srtt has 3 bits to the right of the binary point, 470 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the 471 * binary point, and is smoothed with an ALPHA of 0.75. 472 */ 473 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ 474 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ 475 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ 476 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ 477 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ 478 479 /* 480 * The initial retransmission should happen at rtt + 4 * rttvar. 481 * Because of the way we do the smoothing, srtt and rttvar 482 * will each average +1/2 tick of bias. When we compute 483 * the retransmit timer, we want 1/2 tick of rounding and 484 * 1 extra tick because of +-1/2 tick uncertainty in the 485 * firing of the timer. The bias will give us exactly the 486 * 1.5 tick we need. But, because the bias is 487 * statistical, we have to test that we don't drop below 488 * the minimum feasible timer (which is 2 ticks). 489 * This version of the macro adapted from a paper by Lawrence 490 * Brakmo and Larry Peterson which outlines a problem caused 491 * by insufficient precision in the original implementation, 492 * which results in inappropriately large RTO values for very 493 * fast networks. 494 */ 495 #define TCP_REXMTVAL(tp) \ 496 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ 497 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) 498 499 /* 500 * TCP statistics. 501 * Many of these should be kept per connection, 502 * but that's inconvenient at the moment. 503 */ 504 struct tcpstat { 505 uint64_t tcps_connattempt; /* connections initiated */ 506 uint64_t tcps_accepts; /* connections accepted */ 507 uint64_t tcps_connects; /* connections established */ 508 uint64_t tcps_drops; /* connections dropped */ 509 uint64_t tcps_conndrops; /* embryonic connections dropped */ 510 uint64_t tcps_minmssdrops; /* average minmss too low drops */ 511 uint64_t tcps_closed; /* conn. closed (includes drops) */ 512 uint64_t tcps_segstimed; /* segs where we tried to get rtt */ 513 uint64_t tcps_rttupdated; /* times we succeeded */ 514 uint64_t tcps_delack; /* delayed acks sent */ 515 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ 516 uint64_t tcps_rexmttimeo; /* retransmit timeouts */ 517 uint64_t tcps_persisttimeo; /* persist timeouts */ 518 uint64_t tcps_keeptimeo; /* keepalive timeouts */ 519 uint64_t tcps_keepprobe; /* keepalive probes sent */ 520 uint64_t tcps_keepdrops; /* connections dropped in keepalive */ 521 522 uint64_t tcps_sndtotal; /* total packets sent */ 523 uint64_t tcps_sndpack; /* data packets sent */ 524 uint64_t tcps_sndbyte; /* data bytes sent */ 525 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */ 526 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */ 527 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */ 528 uint64_t tcps_sndacks; /* ack-only packets sent */ 529 uint64_t tcps_sndprobe; /* window probes sent */ 530 uint64_t tcps_sndurg; /* packets sent with URG only */ 531 uint64_t tcps_sndwinup; /* window update-only packets sent */ 532 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ 533 534 uint64_t tcps_rcvtotal; /* total packets received */ 535 uint64_t tcps_rcvpack; /* packets received in sequence */ 536 uint64_t tcps_rcvbyte; /* bytes received in sequence */ 537 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */ 538 uint64_t tcps_rcvbadoff; /* packets received with bad offset */ 539 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */ 540 uint64_t tcps_rcvshort; /* packets received too short */ 541 uint64_t tcps_rcvduppack; /* duplicate-only packets received */ 542 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */ 543 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */ 544 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ 545 uint64_t tcps_rcvoopack; /* out-of-order packets received */ 546 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */ 547 uint64_t tcps_rcvpackafterwin; /* packets with data after window */ 548 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */ 549 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */ 550 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */ 551 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */ 552 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */ 553 uint64_t tcps_rcvackpack; /* rcvd ack packets */ 554 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */ 555 uint64_t tcps_rcvwinupd; /* rcvd window update packets */ 556 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */ 557 uint64_t tcps_predack; /* times hdr predict ok for acks */ 558 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */ 559 uint64_t tcps_pcbcachemiss; 560 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */ 561 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */ 562 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */ 563 uint64_t tcps_usedrtt; /* times RTT initialized from route */ 564 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */ 565 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/ 566 uint64_t tcps_persistdrop; /* timeout in persist state */ 567 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */ 568 uint64_t tcps_mturesent; /* resends due to MTU discovery */ 569 uint64_t tcps_listendrop; /* listen queue overflows */ 570 uint64_t tcps_badrst; /* ignored RSTs in the window */ 571 572 uint64_t tcps_sc_added; /* entry added to syncache */ 573 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */ 574 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */ 575 uint64_t tcps_sc_dropped; /* could not reply to packet */ 576 uint64_t tcps_sc_completed; /* successful extraction of entry */ 577 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */ 578 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */ 579 uint64_t tcps_sc_reset; /* RST removed entry from syncache */ 580 uint64_t tcps_sc_stale; /* timed out or listen socket gone */ 581 uint64_t tcps_sc_aborted; /* syncache entry aborted */ 582 uint64_t tcps_sc_badack; /* removed due to bad ACK */ 583 uint64_t tcps_sc_unreach; /* ICMP unreachable received */ 584 uint64_t tcps_sc_zonefail; /* zalloc() failed */ 585 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */ 586 uint64_t tcps_sc_recvcookie; /* SYN cookie received */ 587 588 uint64_t tcps_hc_added; /* entry added to hostcache */ 589 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */ 590 591 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ 592 593 /* SACK related stats */ 594 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */ 595 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */ 596 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ 597 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ 598 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ 599 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */ 600 601 /* ECN related stats */ 602 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */ 603 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */ 604 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */ 605 uint64_t tcps_ecn_shs; /* ECN successful handshakes */ 606 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */ 607 608 /* TCP_SIGNATURE related stats */ 609 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */ 610 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */ 611 uint64_t tcps_sig_err_buildsig; /* Mismatching signature received */ 612 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */ 613 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */ 614 615 uint64_t _pad[12]; /* 6 UTO, 6 TBD */ 616 }; 617 618 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */ 619 620 #ifdef _KERNEL 621 #define TI_UNLOCKED 1 622 #define TI_RLOCKED 2 623 #include <sys/counter.h> 624 625 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */ 626 /* 627 * In-kernel consumers can use these accessor macros directly to update 628 * stats. 629 */ 630 #define TCPSTAT_ADD(name, val) \ 631 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)) 632 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1) 633 634 /* 635 * Kernel module consumers must use this accessor macro. 636 */ 637 void kmod_tcpstat_inc(int statnum); 638 #define KMOD_TCPSTAT_INC(name) \ 639 kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t)) 640 641 /* 642 * Running TCP connection count by state. 643 */ 644 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]); 645 #define V_tcps_states VNET(tcps_states) 646 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1) 647 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1) 648 649 /* 650 * TCP specific helper hook point identifiers. 651 */ 652 #define HHOOK_TCP_EST_IN 0 653 #define HHOOK_TCP_EST_OUT 1 654 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT 655 656 struct tcp_hhook_data { 657 struct tcpcb *tp; 658 struct tcphdr *th; 659 struct tcpopt *to; 660 long len; 661 int tso; 662 tcp_seq curack; 663 }; 664 #endif 665 666 /* 667 * TCB structure exported to user-land via sysctl(3). 668 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been 669 * included. Not all of our clients do. 670 */ 671 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) 672 struct xtcp_timer { 673 int tt_rexmt; /* retransmit timer */ 674 int tt_persist; /* retransmit persistence */ 675 int tt_keep; /* keepalive */ 676 int tt_2msl; /* 2*msl TIME_WAIT timer */ 677 int tt_delack; /* delayed ACK timer */ 678 int t_rcvtime; /* Time since last packet received */ 679 }; 680 struct xtcpcb { 681 size_t xt_len; 682 struct inpcb xt_inp; 683 struct tcpcb xt_tp; 684 struct xsocket xt_socket; 685 struct xtcp_timer xt_timer; 686 u_quad_t xt_alignment_hack; 687 }; 688 #endif 689 690 /* 691 * Identifiers for TCP sysctl nodes 692 */ 693 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ 694 #define TCPCTL_MSSDFLT 3 /* MSS default */ 695 #define TCPCTL_STATS 4 /* statistics */ 696 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */ 697 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ 698 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ 699 #define TCPCTL_SENDSPACE 8 /* send buffer space */ 700 #define TCPCTL_RECVSPACE 9 /* receive buffer space */ 701 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ 702 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ 703 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ 704 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ 705 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ 706 #define TCPCTL_DROP 15 /* drop tcp connection */ 707 #define TCPCTL_STATES 16 /* connection counts by TCP state */ 708 709 #ifdef _KERNEL 710 #ifdef SYSCTL_DECL 711 SYSCTL_DECL(_net_inet_tcp); 712 SYSCTL_DECL(_net_inet_tcp_sack); 713 MALLOC_DECLARE(M_TCPLOG); 714 #endif 715 716 VNET_DECLARE(struct inpcbhead, tcb); /* queue of active tcpcb's */ 717 VNET_DECLARE(struct inpcbinfo, tcbinfo); 718 extern int tcp_log_in_vain; 719 VNET_DECLARE(int, tcp_mssdflt); /* XXX */ 720 VNET_DECLARE(int, tcp_minmss); 721 VNET_DECLARE(int, tcp_delack_enabled); 722 VNET_DECLARE(int, tcp_do_rfc3390); 723 VNET_DECLARE(int, tcp_initcwnd_segments); 724 VNET_DECLARE(int, tcp_sendspace); 725 VNET_DECLARE(int, tcp_recvspace); 726 VNET_DECLARE(int, path_mtu_discovery); 727 VNET_DECLARE(int, tcp_do_rfc3465); 728 VNET_DECLARE(int, tcp_abc_l_var); 729 #define V_tcb VNET(tcb) 730 #define V_tcbinfo VNET(tcbinfo) 731 #define V_tcp_mssdflt VNET(tcp_mssdflt) 732 #define V_tcp_minmss VNET(tcp_minmss) 733 #define V_tcp_delack_enabled VNET(tcp_delack_enabled) 734 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390) 735 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments) 736 #define V_tcp_sendspace VNET(tcp_sendspace) 737 #define V_tcp_recvspace VNET(tcp_recvspace) 738 #define V_path_mtu_discovery VNET(path_mtu_discovery) 739 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465) 740 #define V_tcp_abc_l_var VNET(tcp_abc_l_var) 741 742 VNET_DECLARE(int, tcp_do_sack); /* SACK enabled/disabled */ 743 VNET_DECLARE(int, tcp_sc_rst_sock_fail); /* RST on sock alloc failure */ 744 #define V_tcp_do_sack VNET(tcp_do_sack) 745 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail) 746 747 VNET_DECLARE(int, tcp_do_ecn); /* TCP ECN enabled/disabled */ 748 VNET_DECLARE(int, tcp_ecn_maxretries); 749 #define V_tcp_do_ecn VNET(tcp_do_ecn) 750 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries) 751 752 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]); 753 #define V_tcp_hhh VNET(tcp_hhh) 754 755 VNET_DECLARE(int, tcp_do_rfc6675_pipe); 756 #define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe) 757 758 int tcp_addoptions(struct tcpopt *, u_char *); 759 int tcp_ccalgounload(struct cc_algo *unload_algo); 760 struct tcpcb * 761 tcp_close(struct tcpcb *); 762 void tcp_discardcb(struct tcpcb *); 763 void tcp_twstart(struct tcpcb *); 764 void tcp_twclose(struct tcptw *, int); 765 void tcp_ctlinput(int, struct sockaddr *, void *); 766 int tcp_ctloutput(struct socket *, struct sockopt *); 767 struct tcpcb * 768 tcp_drop(struct tcpcb *, int); 769 void tcp_drain(void); 770 void tcp_init(void); 771 void tcp_fini(void *); 772 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *, 773 const void *); 774 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *, 775 const void *); 776 int tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *); 777 void tcp_reass_global_init(void); 778 void tcp_reass_flush(struct tcpcb *); 779 void tcp_dooptions(struct tcpopt *, u_char *, int, int); 780 void tcp_dropwithreset(struct mbuf *, struct tcphdr *, 781 struct tcpcb *, int, int); 782 void tcp_pulloutofband(struct socket *, 783 struct tcphdr *, struct mbuf *, int); 784 void tcp_xmit_timer(struct tcpcb *, int); 785 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); 786 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th, 787 uint16_t nsegs, uint16_t type); 788 void cc_conn_init(struct tcpcb *tp); 789 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); 790 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type); 791 void hhook_run_tcp_est_in(struct tcpcb *tp, 792 struct tcphdr *th, struct tcpopt *to); 793 794 int tcp_input(struct mbuf **, int *, int); 795 void tcp_do_segment(struct mbuf *, struct tcphdr *, 796 struct socket *, struct tcpcb *, int, int, uint8_t, 797 int); 798 799 int register_tcp_functions(struct tcp_function_block *blk, int wait); 800 int deregister_tcp_functions(struct tcp_function_block *blk); 801 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs); 802 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk); 803 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp); 804 805 u_long tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *); 806 u_long tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *); 807 u_int tcp_maxseg(const struct tcpcb *); 808 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *, 809 struct tcp_ifcap *); 810 void tcp_mss(struct tcpcb *, int); 811 int tcp_mssopt(struct in_conninfo *); 812 struct inpcb * 813 tcp_drop_syn_sent(struct inpcb *, int); 814 struct tcpcb * 815 tcp_newtcpcb(struct inpcb *); 816 int tcp_output(struct tcpcb *); 817 void tcp_state_change(struct tcpcb *, int); 818 void tcp_respond(struct tcpcb *, void *, 819 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int); 820 void tcp_tw_init(void); 821 #ifdef VIMAGE 822 void tcp_tw_destroy(void); 823 #endif 824 void tcp_tw_zone_change(void); 825 int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *, 826 struct mbuf *, int); 827 void tcp_setpersist(struct tcpcb *); 828 #ifdef TCP_SIGNATURE 829 struct secasvar; 830 struct secasvar *tcp_get_sav(struct mbuf *, u_int); 831 int tcp_signature_do_compute(struct mbuf *, int, int, u_char *, 832 struct secasvar *); 833 int tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int); 834 int tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *, 835 struct tcphdr *, u_int); 836 int tcp_signature_check(struct mbuf *m, int off0, int tlen, int optlen, 837 struct tcpopt *to, struct tcphdr *th, u_int tcpbflag); 838 #endif 839 void tcp_slowtimo(void); 840 struct tcptemp * 841 tcpip_maketemplate(struct inpcb *); 842 void tcpip_fillheaders(struct inpcb *, void *, void *); 843 void tcp_timer_activate(struct tcpcb *, uint32_t, u_int); 844 int tcp_timer_active(struct tcpcb *, uint32_t); 845 void tcp_timer_stop(struct tcpcb *, uint32_t); 846 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int); 847 /* 848 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) 849 */ 850 void tcp_hc_init(void); 851 #ifdef VIMAGE 852 void tcp_hc_destroy(void); 853 #endif 854 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); 855 u_long tcp_hc_getmtu(struct in_conninfo *); 856 void tcp_hc_updatemtu(struct in_conninfo *, u_long); 857 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); 858 859 extern struct pr_usrreqs tcp_usrreqs; 860 tcp_seq tcp_new_isn(struct tcpcb *); 861 862 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); 863 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); 864 void tcp_clean_sackreport(struct tcpcb *tp); 865 void tcp_sack_adjust(struct tcpcb *tp); 866 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); 867 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); 868 void tcp_free_sackholes(struct tcpcb *tp); 869 int tcp_newreno(struct tcpcb *, struct tcphdr *); 870 u_long tcp_seq_subtract(u_long, u_long ); 871 int tcp_compute_pipe(struct tcpcb *); 872 873 static inline void 874 tcp_fields_to_host(struct tcphdr *th) 875 { 876 877 th->th_seq = ntohl(th->th_seq); 878 th->th_ack = ntohl(th->th_ack); 879 th->th_win = ntohs(th->th_win); 880 th->th_urp = ntohs(th->th_urp); 881 } 882 883 #ifdef TCP_SIGNATURE 884 static inline void 885 tcp_fields_to_net(struct tcphdr *th) 886 { 887 888 th->th_seq = htonl(th->th_seq); 889 th->th_ack = htonl(th->th_ack); 890 th->th_win = htons(th->th_win); 891 th->th_urp = htons(th->th_urp); 892 } 893 #endif 894 #endif /* _KERNEL */ 895 896 #endif /* _NETINET_TCP_VAR_H_ */ 897