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