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 uint32_t rcv_wnd; /* receive window */ 196 tcp_seq rcv_up; /* receive urgent pointer */ 197 198 uint32_t snd_wnd; /* send window */ 199 uint32_t snd_cwnd; /* congestion-controlled window */ 200 u_long snd_spare1; /* unused */ 201 uint32_t 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 uint32_t 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 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */ 243 uint32_t 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 #if defined(_KERNEL) && !defined(TCP_RFC7413) 353 #define IS_FASTOPEN(t_flags) (false) 354 #else 355 #define IS_FASTOPEN(t_flags) (t_flags & TF_FASTOPEN) 356 #endif 357 358 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una) 359 360 /* 361 * Flags for the t_oobflags field. 362 */ 363 #define TCPOOB_HAVEDATA 0x01 364 #define TCPOOB_HADDATA 0x02 365 366 #ifdef TCP_SIGNATURE 367 /* 368 * Defines which are needed by the xform_tcp module and tcp_[in|out]put 369 * for SADB verification and lookup. 370 */ 371 #define TCP_SIGLEN 16 /* length of computed digest in bytes */ 372 #define TCP_KEYLEN_MIN 1 /* minimum length of TCP-MD5 key */ 373 #define TCP_KEYLEN_MAX 80 /* maximum length of TCP-MD5 key */ 374 /* 375 * Only a single SA per host may be specified at this time. An SPI is 376 * needed in order for the KEY_ALLOCSA() lookup to work. 377 */ 378 #define TCP_SIG_SPI 0x1000 379 #endif /* TCP_SIGNATURE */ 380 381 /* 382 * Flags for PLPMTU handling, t_flags2 383 */ 384 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */ 385 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */ 386 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */ 387 388 /* 389 * Structure to hold TCP options that are only used during segment 390 * processing (in tcp_input), but not held in the tcpcb. 391 * It's basically used to reduce the number of parameters 392 * to tcp_dooptions and tcp_addoptions. 393 * The binary order of the to_flags is relevant for packing of the 394 * options in tcp_addoptions. 395 */ 396 struct tcpopt { 397 u_int32_t to_flags; /* which options are present */ 398 #define TOF_MSS 0x0001 /* maximum segment size */ 399 #define TOF_SCALE 0x0002 /* window scaling */ 400 #define TOF_SACKPERM 0x0004 /* SACK permitted */ 401 #define TOF_TS 0x0010 /* timestamp */ 402 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */ 403 #define TOF_SACK 0x0080 /* Peer sent SACK option */ 404 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */ 405 #define TOF_MAXOPT 0x0200 406 u_int32_t to_tsval; /* new timestamp */ 407 u_int32_t to_tsecr; /* reflected timestamp */ 408 u_char *to_sacks; /* pointer to the first SACK blocks */ 409 u_char *to_signature; /* pointer to the TCP-MD5 signature */ 410 u_char *to_tfo_cookie; /* pointer to the TFO cookie */ 411 u_int16_t to_mss; /* maximum segment size */ 412 u_int8_t to_wscale; /* window scaling */ 413 u_int8_t to_nsacks; /* number of SACK blocks */ 414 u_int8_t to_tfo_len; /* TFO cookie length */ 415 u_int32_t to_spare; /* UTO */ 416 }; 417 418 /* 419 * Flags for tcp_dooptions. 420 */ 421 #define TO_SYN 0x01 /* parse SYN-only options */ 422 423 struct hc_metrics_lite { /* must stay in sync with hc_metrics */ 424 uint32_t rmx_mtu; /* MTU for this path */ 425 uint32_t rmx_ssthresh; /* outbound gateway buffer limit */ 426 uint32_t rmx_rtt; /* estimated round trip time */ 427 uint32_t rmx_rttvar; /* estimated rtt variance */ 428 uint32_t rmx_cwnd; /* congestion window */ 429 uint32_t rmx_sendpipe; /* outbound delay-bandwidth product */ 430 uint32_t rmx_recvpipe; /* inbound delay-bandwidth product */ 431 }; 432 433 /* 434 * Used by tcp_maxmtu() to communicate interface specific features 435 * and limits at the time of connection setup. 436 */ 437 struct tcp_ifcap { 438 int ifcap; 439 u_int tsomax; 440 u_int tsomaxsegcount; 441 u_int tsomaxsegsize; 442 }; 443 444 #ifndef _NETINET_IN_PCB_H_ 445 struct in_conninfo; 446 #endif /* _NETINET_IN_PCB_H_ */ 447 448 struct tcptw { 449 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */ 450 tcp_seq snd_nxt; 451 tcp_seq rcv_nxt; 452 tcp_seq iss; 453 tcp_seq irs; 454 u_short last_win; /* cached window value */ 455 u_short tw_so_options; /* copy of so_options */ 456 struct ucred *tw_cred; /* user credentials */ 457 u_int32_t t_recent; 458 u_int32_t ts_offset; /* our timestamp offset */ 459 u_int t_starttime; 460 int tw_time; 461 TAILQ_ENTRY(tcptw) tw_2msl; 462 void *tw_pspare; /* TCP_SIGNATURE */ 463 u_int *tw_spare; /* TCP_SIGNATURE */ 464 }; 465 466 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) 467 #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb) 468 #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) 469 470 /* 471 * The smoothed round-trip time and estimated variance 472 * are stored as fixed point numbers scaled by the values below. 473 * For convenience, these scales are also used in smoothing the average 474 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). 475 * With these scales, srtt has 3 bits to the right of the binary point, 476 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the 477 * binary point, and is smoothed with an ALPHA of 0.75. 478 */ 479 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ 480 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ 481 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ 482 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ 483 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ 484 485 /* 486 * The initial retransmission should happen at rtt + 4 * rttvar. 487 * Because of the way we do the smoothing, srtt and rttvar 488 * will each average +1/2 tick of bias. When we compute 489 * the retransmit timer, we want 1/2 tick of rounding and 490 * 1 extra tick because of +-1/2 tick uncertainty in the 491 * firing of the timer. The bias will give us exactly the 492 * 1.5 tick we need. But, because the bias is 493 * statistical, we have to test that we don't drop below 494 * the minimum feasible timer (which is 2 ticks). 495 * This version of the macro adapted from a paper by Lawrence 496 * Brakmo and Larry Peterson which outlines a problem caused 497 * by insufficient precision in the original implementation, 498 * which results in inappropriately large RTO values for very 499 * fast networks. 500 */ 501 #define TCP_REXMTVAL(tp) \ 502 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ 503 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) 504 505 /* 506 * TCP statistics. 507 * Many of these should be kept per connection, 508 * but that's inconvenient at the moment. 509 */ 510 struct tcpstat { 511 uint64_t tcps_connattempt; /* connections initiated */ 512 uint64_t tcps_accepts; /* connections accepted */ 513 uint64_t tcps_connects; /* connections established */ 514 uint64_t tcps_drops; /* connections dropped */ 515 uint64_t tcps_conndrops; /* embryonic connections dropped */ 516 uint64_t tcps_minmssdrops; /* average minmss too low drops */ 517 uint64_t tcps_closed; /* conn. closed (includes drops) */ 518 uint64_t tcps_segstimed; /* segs where we tried to get rtt */ 519 uint64_t tcps_rttupdated; /* times we succeeded */ 520 uint64_t tcps_delack; /* delayed acks sent */ 521 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ 522 uint64_t tcps_rexmttimeo; /* retransmit timeouts */ 523 uint64_t tcps_persisttimeo; /* persist timeouts */ 524 uint64_t tcps_keeptimeo; /* keepalive timeouts */ 525 uint64_t tcps_keepprobe; /* keepalive probes sent */ 526 uint64_t tcps_keepdrops; /* connections dropped in keepalive */ 527 528 uint64_t tcps_sndtotal; /* total packets sent */ 529 uint64_t tcps_sndpack; /* data packets sent */ 530 uint64_t tcps_sndbyte; /* data bytes sent */ 531 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */ 532 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */ 533 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */ 534 uint64_t tcps_sndacks; /* ack-only packets sent */ 535 uint64_t tcps_sndprobe; /* window probes sent */ 536 uint64_t tcps_sndurg; /* packets sent with URG only */ 537 uint64_t tcps_sndwinup; /* window update-only packets sent */ 538 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ 539 540 uint64_t tcps_rcvtotal; /* total packets received */ 541 uint64_t tcps_rcvpack; /* packets received in sequence */ 542 uint64_t tcps_rcvbyte; /* bytes received in sequence */ 543 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */ 544 uint64_t tcps_rcvbadoff; /* packets received with bad offset */ 545 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */ 546 uint64_t tcps_rcvshort; /* packets received too short */ 547 uint64_t tcps_rcvduppack; /* duplicate-only packets received */ 548 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */ 549 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */ 550 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ 551 uint64_t tcps_rcvoopack; /* out-of-order packets received */ 552 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */ 553 uint64_t tcps_rcvpackafterwin; /* packets with data after window */ 554 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */ 555 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */ 556 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */ 557 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */ 558 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */ 559 uint64_t tcps_rcvackpack; /* rcvd ack packets */ 560 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */ 561 uint64_t tcps_rcvwinupd; /* rcvd window update packets */ 562 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */ 563 uint64_t tcps_predack; /* times hdr predict ok for acks */ 564 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */ 565 uint64_t tcps_pcbcachemiss; 566 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */ 567 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */ 568 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */ 569 uint64_t tcps_usedrtt; /* times RTT initialized from route */ 570 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */ 571 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/ 572 uint64_t tcps_persistdrop; /* timeout in persist state */ 573 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */ 574 uint64_t tcps_mturesent; /* resends due to MTU discovery */ 575 uint64_t tcps_listendrop; /* listen queue overflows */ 576 uint64_t tcps_badrst; /* ignored RSTs in the window */ 577 578 uint64_t tcps_sc_added; /* entry added to syncache */ 579 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */ 580 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */ 581 uint64_t tcps_sc_dropped; /* could not reply to packet */ 582 uint64_t tcps_sc_completed; /* successful extraction of entry */ 583 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */ 584 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */ 585 uint64_t tcps_sc_reset; /* RST removed entry from syncache */ 586 uint64_t tcps_sc_stale; /* timed out or listen socket gone */ 587 uint64_t tcps_sc_aborted; /* syncache entry aborted */ 588 uint64_t tcps_sc_badack; /* removed due to bad ACK */ 589 uint64_t tcps_sc_unreach; /* ICMP unreachable received */ 590 uint64_t tcps_sc_zonefail; /* zalloc() failed */ 591 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */ 592 uint64_t tcps_sc_recvcookie; /* SYN cookie received */ 593 594 uint64_t tcps_hc_added; /* entry added to hostcache */ 595 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */ 596 597 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ 598 599 /* SACK related stats */ 600 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */ 601 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */ 602 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ 603 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ 604 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ 605 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */ 606 607 /* ECN related stats */ 608 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */ 609 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */ 610 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */ 611 uint64_t tcps_ecn_shs; /* ECN successful handshakes */ 612 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */ 613 614 /* TCP_SIGNATURE related stats */ 615 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */ 616 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */ 617 uint64_t tcps_sig_err_buildsig; /* Mismatching signature received */ 618 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */ 619 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */ 620 621 uint64_t _pad[12]; /* 6 UTO, 6 TBD */ 622 }; 623 624 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */ 625 626 #ifdef _KERNEL 627 #define TI_UNLOCKED 1 628 #define TI_RLOCKED 2 629 #include <sys/counter.h> 630 631 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */ 632 /* 633 * In-kernel consumers can use these accessor macros directly to update 634 * stats. 635 */ 636 #define TCPSTAT_ADD(name, val) \ 637 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)) 638 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1) 639 640 /* 641 * Kernel module consumers must use this accessor macro. 642 */ 643 void kmod_tcpstat_inc(int statnum); 644 #define KMOD_TCPSTAT_INC(name) \ 645 kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t)) 646 647 /* 648 * Running TCP connection count by state. 649 */ 650 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]); 651 #define V_tcps_states VNET(tcps_states) 652 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1) 653 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1) 654 655 /* 656 * TCP specific helper hook point identifiers. 657 */ 658 #define HHOOK_TCP_EST_IN 0 659 #define HHOOK_TCP_EST_OUT 1 660 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT 661 662 struct tcp_hhook_data { 663 struct tcpcb *tp; 664 struct tcphdr *th; 665 struct tcpopt *to; 666 uint32_t len; 667 int tso; 668 tcp_seq curack; 669 }; 670 #endif 671 672 /* 673 * TCB structure exported to user-land via sysctl(3). 674 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been 675 * included. Not all of our clients do. 676 */ 677 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) 678 struct xtcp_timer { 679 int tt_rexmt; /* retransmit timer */ 680 int tt_persist; /* retransmit persistence */ 681 int tt_keep; /* keepalive */ 682 int tt_2msl; /* 2*msl TIME_WAIT timer */ 683 int tt_delack; /* delayed ACK timer */ 684 int t_rcvtime; /* Time since last packet received */ 685 }; 686 struct xtcpcb { 687 size_t xt_len; 688 struct inpcb xt_inp; 689 struct tcpcb xt_tp; 690 struct xsocket xt_socket; 691 struct xtcp_timer xt_timer; 692 u_quad_t xt_alignment_hack; 693 }; 694 #endif 695 696 /* 697 * Identifiers for TCP sysctl nodes 698 */ 699 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ 700 #define TCPCTL_MSSDFLT 3 /* MSS default */ 701 #define TCPCTL_STATS 4 /* statistics */ 702 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */ 703 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ 704 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ 705 #define TCPCTL_SENDSPACE 8 /* send buffer space */ 706 #define TCPCTL_RECVSPACE 9 /* receive buffer space */ 707 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ 708 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ 709 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ 710 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ 711 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ 712 #define TCPCTL_DROP 15 /* drop tcp connection */ 713 #define TCPCTL_STATES 16 /* connection counts by TCP state */ 714 715 #ifdef _KERNEL 716 #ifdef SYSCTL_DECL 717 SYSCTL_DECL(_net_inet_tcp); 718 SYSCTL_DECL(_net_inet_tcp_sack); 719 MALLOC_DECLARE(M_TCPLOG); 720 #endif 721 722 VNET_DECLARE(struct inpcbhead, tcb); /* queue of active tcpcb's */ 723 VNET_DECLARE(struct inpcbinfo, tcbinfo); 724 extern int tcp_log_in_vain; 725 VNET_DECLARE(int, tcp_mssdflt); /* XXX */ 726 VNET_DECLARE(int, tcp_minmss); 727 VNET_DECLARE(int, tcp_delack_enabled); 728 VNET_DECLARE(int, tcp_do_rfc3390); 729 VNET_DECLARE(int, tcp_initcwnd_segments); 730 VNET_DECLARE(int, tcp_sendspace); 731 VNET_DECLARE(int, tcp_recvspace); 732 VNET_DECLARE(int, path_mtu_discovery); 733 VNET_DECLARE(int, tcp_do_rfc3465); 734 VNET_DECLARE(int, tcp_abc_l_var); 735 #define V_tcb VNET(tcb) 736 #define V_tcbinfo VNET(tcbinfo) 737 #define V_tcp_mssdflt VNET(tcp_mssdflt) 738 #define V_tcp_minmss VNET(tcp_minmss) 739 #define V_tcp_delack_enabled VNET(tcp_delack_enabled) 740 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390) 741 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments) 742 #define V_tcp_sendspace VNET(tcp_sendspace) 743 #define V_tcp_recvspace VNET(tcp_recvspace) 744 #define V_path_mtu_discovery VNET(path_mtu_discovery) 745 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465) 746 #define V_tcp_abc_l_var VNET(tcp_abc_l_var) 747 748 VNET_DECLARE(int, tcp_do_sack); /* SACK enabled/disabled */ 749 VNET_DECLARE(int, tcp_sc_rst_sock_fail); /* RST on sock alloc failure */ 750 #define V_tcp_do_sack VNET(tcp_do_sack) 751 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail) 752 753 VNET_DECLARE(int, tcp_do_ecn); /* TCP ECN enabled/disabled */ 754 VNET_DECLARE(int, tcp_ecn_maxretries); 755 #define V_tcp_do_ecn VNET(tcp_do_ecn) 756 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries) 757 758 #ifdef TCP_HHOOK 759 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]); 760 #define V_tcp_hhh VNET(tcp_hhh) 761 #endif 762 763 VNET_DECLARE(int, tcp_do_rfc6675_pipe); 764 #define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe) 765 766 int tcp_addoptions(struct tcpopt *, u_char *); 767 int tcp_ccalgounload(struct cc_algo *unload_algo); 768 struct tcpcb * 769 tcp_close(struct tcpcb *); 770 void tcp_discardcb(struct tcpcb *); 771 void tcp_twstart(struct tcpcb *); 772 void tcp_twclose(struct tcptw *, int); 773 void tcp_ctlinput(int, struct sockaddr *, void *); 774 int tcp_ctloutput(struct socket *, struct sockopt *); 775 struct tcpcb * 776 tcp_drop(struct tcpcb *, int); 777 void tcp_drain(void); 778 void tcp_init(void); 779 void tcp_fini(void *); 780 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *, 781 const void *); 782 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *, 783 const void *); 784 int tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *); 785 void tcp_reass_global_init(void); 786 void tcp_reass_flush(struct tcpcb *); 787 void tcp_dooptions(struct tcpopt *, u_char *, int, int); 788 void tcp_dropwithreset(struct mbuf *, struct tcphdr *, 789 struct tcpcb *, int, int); 790 void tcp_pulloutofband(struct socket *, 791 struct tcphdr *, struct mbuf *, int); 792 void tcp_xmit_timer(struct tcpcb *, int); 793 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); 794 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th, 795 uint16_t nsegs, uint16_t type); 796 void cc_conn_init(struct tcpcb *tp); 797 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); 798 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type); 799 #ifdef TCP_HHOOK 800 void hhook_run_tcp_est_in(struct tcpcb *tp, 801 struct tcphdr *th, struct tcpopt *to); 802 #endif 803 804 int tcp_input(struct mbuf **, int *, int); 805 void tcp_do_segment(struct mbuf *, struct tcphdr *, 806 struct socket *, struct tcpcb *, int, int, uint8_t, 807 int); 808 809 int register_tcp_functions(struct tcp_function_block *blk, int wait); 810 int deregister_tcp_functions(struct tcp_function_block *blk); 811 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs); 812 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk); 813 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp); 814 815 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *); 816 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *); 817 u_int tcp_maxseg(const struct tcpcb *); 818 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *, 819 struct tcp_ifcap *); 820 void tcp_mss(struct tcpcb *, int); 821 int tcp_mssopt(struct in_conninfo *); 822 struct inpcb * 823 tcp_drop_syn_sent(struct inpcb *, int); 824 struct tcpcb * 825 tcp_newtcpcb(struct inpcb *); 826 int tcp_output(struct tcpcb *); 827 void tcp_state_change(struct tcpcb *, int); 828 void tcp_respond(struct tcpcb *, void *, 829 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int); 830 void tcp_tw_init(void); 831 #ifdef VIMAGE 832 void tcp_tw_destroy(void); 833 #endif 834 void tcp_tw_zone_change(void); 835 int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *, 836 struct mbuf *, int); 837 void tcp_setpersist(struct tcpcb *); 838 #ifdef TCP_SIGNATURE 839 struct secasvar; 840 struct secasvar *tcp_get_sav(struct mbuf *, u_int); 841 int tcp_signature_do_compute(struct mbuf *, int, int, u_char *, 842 struct secasvar *); 843 int tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int); 844 int tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *, 845 struct tcphdr *, u_int); 846 int tcp_signature_check(struct mbuf *m, int off0, int tlen, int optlen, 847 struct tcpopt *to, struct tcphdr *th, u_int tcpbflag); 848 #endif 849 void tcp_slowtimo(void); 850 struct tcptemp * 851 tcpip_maketemplate(struct inpcb *); 852 void tcpip_fillheaders(struct inpcb *, void *, void *); 853 void tcp_timer_activate(struct tcpcb *, uint32_t, u_int); 854 int tcp_timer_active(struct tcpcb *, uint32_t); 855 void tcp_timer_stop(struct tcpcb *, uint32_t); 856 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int); 857 /* 858 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) 859 */ 860 void tcp_hc_init(void); 861 #ifdef VIMAGE 862 void tcp_hc_destroy(void); 863 #endif 864 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); 865 uint32_t tcp_hc_getmtu(struct in_conninfo *); 866 void tcp_hc_updatemtu(struct in_conninfo *, uint32_t); 867 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); 868 869 extern struct pr_usrreqs tcp_usrreqs; 870 tcp_seq tcp_new_isn(struct tcpcb *); 871 872 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); 873 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); 874 void tcp_clean_sackreport(struct tcpcb *tp); 875 void tcp_sack_adjust(struct tcpcb *tp); 876 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); 877 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); 878 void tcp_free_sackholes(struct tcpcb *tp); 879 int tcp_newreno(struct tcpcb *, struct tcphdr *); 880 int tcp_compute_pipe(struct tcpcb *); 881 882 static inline void 883 tcp_fields_to_host(struct tcphdr *th) 884 { 885 886 th->th_seq = ntohl(th->th_seq); 887 th->th_ack = ntohl(th->th_ack); 888 th->th_win = ntohs(th->th_win); 889 th->th_urp = ntohs(th->th_urp); 890 } 891 892 #ifdef TCP_SIGNATURE 893 static inline void 894 tcp_fields_to_net(struct tcphdr *th) 895 { 896 897 th->th_seq = htonl(th->th_seq); 898 th->th_ack = htonl(th->th_ack); 899 th->th_win = htons(th->th_win); 900 th->th_urp = htons(th->th_urp); 901 } 902 #endif 903 #endif /* _KERNEL */ 904 905 #endif /* _NETINET_TCP_VAR_H_ */ 906