1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1993, 1994, 1995 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95 32 * $FreeBSD$ 33 */ 34 35 #ifndef _NETINET_TCP_VAR_H_ 36 #define _NETINET_TCP_VAR_H_ 37 38 #include <netinet/tcp.h> 39 #include <netinet/tcp_fsm.h> 40 41 #ifdef _KERNEL 42 #include <net/vnet.h> 43 #include <sys/mbuf.h> 44 #endif 45 46 #define TCP_END_BYTE_INFO 8 /* Bytes that makeup the "end information array" */ 47 /* Types of ending byte info */ 48 #define TCP_EI_EMPTY_SLOT 0 49 #define TCP_EI_STATUS_CLIENT_FIN 0x1 50 #define TCP_EI_STATUS_CLIENT_RST 0x2 51 #define TCP_EI_STATUS_SERVER_FIN 0x3 52 #define TCP_EI_STATUS_SERVER_RST 0x4 53 #define TCP_EI_STATUS_RETRAN 0x5 54 #define TCP_EI_STATUS_PROGRESS 0x6 55 #define TCP_EI_STATUS_PERSIST_MAX 0x7 56 #define TCP_EI_STATUS_KEEP_MAX 0x8 57 #define TCP_EI_STATUS_DATA_A_CLOSE 0x9 58 #define TCP_EI_STATUS_RST_IN_FRONT 0xa 59 #define TCP_EI_STATUS_2MSL 0xb 60 #define TCP_EI_STATUS_MAX_VALUE 0xb 61 62 /************************************************/ 63 /* Status bits we track to assure no duplicates, 64 * the bits here are not used by the code but 65 * for human representation. To check a bit we 66 * take and shift over by 1 minus the value (1-8). 67 */ 68 /************************************************/ 69 #define TCP_EI_BITS_CLIENT_FIN 0x001 70 #define TCP_EI_BITS_CLIENT_RST 0x002 71 #define TCP_EI_BITS_SERVER_FIN 0x004 72 #define TCP_EI_BITS_SERVER_RST 0x008 73 #define TCP_EI_BITS_RETRAN 0x010 74 #define TCP_EI_BITS_PROGRESS 0x020 75 #define TCP_EI_BITS_PRESIST_MAX 0x040 76 #define TCP_EI_BITS_KEEP_MAX 0x080 77 #define TCP_EI_BITS_DATA_A_CLO 0x100 78 #define TCP_EI_BITS_RST_IN_FR 0x200 /* a front state reset */ 79 #define TCP_EI_BITS_2MS_TIMER 0x400 /* 2 MSL timer expired */ 80 81 #if defined(_KERNEL) || defined(_WANT_TCPCB) 82 /* TCP segment queue entry */ 83 struct tseg_qent { 84 TAILQ_ENTRY(tseg_qent) tqe_q; 85 struct mbuf *tqe_m; /* mbuf contains packet */ 86 struct mbuf *tqe_last; /* last mbuf in chain */ 87 tcp_seq tqe_start; /* TCP Sequence number start */ 88 int tqe_len; /* TCP segment data length */ 89 uint32_t tqe_flags; /* The flags from the th->th_flags */ 90 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */ 91 }; 92 TAILQ_HEAD(tsegqe_head, tseg_qent); 93 94 struct sackblk { 95 tcp_seq start; /* start seq no. of sack block */ 96 tcp_seq end; /* end seq no. */ 97 }; 98 99 struct sackhole { 100 tcp_seq start; /* start seq no. of hole */ 101 tcp_seq end; /* end seq no. */ 102 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ 103 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ 104 }; 105 106 struct sackhint { 107 struct sackhole *nexthole; 108 int32_t sack_bytes_rexmit; 109 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */ 110 111 int32_t delivered_data; /* Newly acked data from last SACK */ 112 113 int32_t sacked_bytes; /* Total sacked bytes reported by the 114 * receiver via sack option 115 */ 116 uint32_t recover_fs; /* Flight Size at the start of Loss recovery */ 117 uint32_t prr_delivered; /* Total bytes delivered using PRR */ 118 uint32_t prr_out; /* Bytes sent during IN_RECOVERY */ 119 }; 120 121 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq) 122 123 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem); 124 125 /* 126 * Tcp control block, one per tcp; fields: 127 * Organized for 64 byte cacheline efficiency based 128 * on common tcp_input/tcp_output processing. 129 */ 130 struct tcpcb { 131 /* Cache line 1 */ 132 struct inpcb *t_inpcb; /* back pointer to internet pcb */ 133 struct tcp_function_block *t_fb;/* TCP function call block */ 134 void *t_fb_ptr; /* Pointer to t_fb specific data */ 135 uint32_t t_maxseg:24, /* maximum segment size */ 136 t_logstate:8; /* State of "black box" logging */ 137 uint32_t t_port:16, /* Tunneling (over udp) port */ 138 t_state:4, /* state of this connection */ 139 t_idle_reduce : 1, 140 t_delayed_ack: 7, /* Delayed ack variable */ 141 t_fin_is_rst: 1, /* Are fin's treated as resets */ 142 t_log_state_set: 1, 143 bits_spare : 2; 144 u_int t_flags; 145 tcp_seq snd_una; /* sent but unacknowledged */ 146 tcp_seq snd_max; /* highest sequence number sent; 147 * used to recognize retransmits 148 */ 149 tcp_seq snd_nxt; /* send next */ 150 tcp_seq snd_up; /* send urgent pointer */ 151 uint32_t snd_wnd; /* send window */ 152 uint32_t snd_cwnd; /* congestion-controlled window */ 153 uint32_t t_peakrate_thr; /* pre-calculated peak rate threshold */ 154 /* Cache line 2 */ 155 u_int32_t ts_offset; /* our timestamp offset */ 156 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ 157 int rcv_numsacks; /* # distinct sack blks present */ 158 u_int t_tsomax; /* TSO total burst length limit in bytes */ 159 u_int t_tsomaxsegcount; /* TSO maximum segment count */ 160 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */ 161 tcp_seq rcv_nxt; /* receive next */ 162 tcp_seq rcv_adv; /* advertised window */ 163 uint32_t rcv_wnd; /* receive window */ 164 u_int t_flags2; /* More tcpcb flags storage */ 165 int t_srtt; /* smoothed round-trip time */ 166 int t_rttvar; /* variance in round-trip time */ 167 u_int32_t ts_recent; /* timestamp echo data */ 168 u_char snd_scale; /* window scaling for send window */ 169 u_char rcv_scale; /* window scaling for recv window */ 170 u_char snd_limited; /* segments limited transmitted */ 171 u_char request_r_scale; /* pending window scaling */ 172 tcp_seq last_ack_sent; 173 u_int t_rcvtime; /* inactivity time */ 174 /* Cache line 3 */ 175 tcp_seq rcv_up; /* receive urgent pointer */ 176 int t_segqlen; /* segment reassembly queue length */ 177 uint32_t t_segqmbuflen; /* Count of bytes mbufs on all entries */ 178 struct tsegqe_head t_segq; /* segment reassembly queue */ 179 struct mbuf *t_in_pkt; 180 struct mbuf *t_tail_pkt; 181 struct tcp_timer *t_timers; /* All the TCP timers in one struct */ 182 struct vnet *t_vnet; /* back pointer to parent vnet */ 183 uint32_t snd_ssthresh; /* snd_cwnd size threshold for 184 * for slow start exponential to 185 * linear switch 186 */ 187 tcp_seq snd_wl1; /* window update seg seq number */ 188 /* Cache line 4 */ 189 tcp_seq snd_wl2; /* window update seg ack number */ 190 191 tcp_seq irs; /* initial receive sequence number */ 192 tcp_seq iss; /* initial send sequence number */ 193 u_int t_acktime; /* RACK and BBR incoming new data was acked */ 194 u_int t_sndtime; /* time last data was sent */ 195 u_int ts_recent_age; /* when last updated */ 196 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ 197 uint16_t cl4_spare; /* Spare to adjust CL 4 */ 198 char t_oobflags; /* have some */ 199 char t_iobc; /* input character */ 200 int t_rxtcur; /* current retransmit value (ticks) */ 201 202 int t_rxtshift; /* log(2) of rexmt exp. backoff */ 203 u_int t_rtttime; /* RTT measurement start time */ 204 205 tcp_seq t_rtseq; /* sequence number being timed */ 206 u_int t_starttime; /* time connection was established */ 207 u_int t_fbyte_in; /* ticks time when first byte queued in */ 208 u_int t_fbyte_out; /* ticks time when first byte queued out */ 209 210 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */ 211 int t_blackhole_enter; /* when to enter blackhole detection */ 212 int t_blackhole_exit; /* when to exit blackhole detection */ 213 u_int t_rttmin; /* minimum rtt allowed */ 214 215 u_int t_rttbest; /* best rtt we've seen */ 216 217 int t_softerror; /* possible error not yet reported */ 218 uint32_t max_sndwnd; /* largest window peer has offered */ 219 /* Cache line 5 */ 220 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */ 221 uint32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */ 222 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ 223 int t_sndzerowin; /* zero-window updates sent */ 224 u_long t_rttupdated; /* number of times rtt sampled */ 225 int snd_numholes; /* number of holes seen by sender */ 226 u_int t_badrxtwin; /* window for retransmit recovery */ 227 TAILQ_HEAD(sackhole_head, sackhole) snd_holes; 228 /* SACK scoreboard (sorted) */ 229 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ 230 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ 231 struct sackhint sackhint; /* SACK scoreboard hint */ 232 int t_rttlow; /* smallest observerved RTT */ 233 int rfbuf_cnt; /* recv buffer autoscaling byte count */ 234 struct toedev *tod; /* toedev handling this connection */ 235 int t_sndrexmitpack; /* retransmit packets sent */ 236 int t_rcvoopack; /* out-of-order packets received */ 237 void *t_toe; /* TOE pcb pointer */ 238 struct cc_algo *cc_algo; /* congestion control algorithm */ 239 struct cc_var *ccv; /* congestion control specific vars */ 240 struct osd *osd; /* storage for Khelp module data */ 241 int t_bytes_acked; /* # bytes acked during current RTT */ 242 u_int t_maxunacktime; 243 u_int t_keepinit; /* time to establish connection */ 244 u_int t_keepidle; /* time before keepalive probes begin */ 245 u_int t_keepintvl; /* interval between keepalives */ 246 u_int t_keepcnt; /* number of keepalives before close */ 247 int t_dupacks; /* consecutive dup acks recd */ 248 int t_lognum; /* Number of log entries */ 249 int t_loglimit; /* Maximum number of log entries */ 250 int64_t t_pacing_rate; /* bytes / sec, -1 => unlimited */ 251 struct tcp_log_stailq t_logs; /* Log buffer */ 252 struct tcp_log_id_node *t_lin; 253 struct tcp_log_id_bucket *t_lib; 254 const char *t_output_caller; /* Function that called tcp_output */ 255 struct statsblob *t_stats; /* Per-connection stats */ 256 uint32_t t_logsn; /* Log "serial number" */ 257 uint32_t gput_ts; /* Time goodput measurement started */ 258 tcp_seq gput_seq; /* Outbound measurement seq */ 259 tcp_seq gput_ack; /* Inbound measurement ack */ 260 int32_t t_stats_gput_prev; /* XXXLAS: Prev gput measurement */ 261 uint32_t t_maxpeakrate; /* max peak rate set by user, in bytes/s */ 262 uint32_t t_sndtlppack; /* tail loss probe packets sent */ 263 uint64_t t_sndtlpbyte; /* total tail loss probe bytes sent */ 264 uint64_t t_sndbytes; /* total bytes sent */ 265 uint64_t t_snd_rxt_bytes; /* total bytes retransmitted */ 266 267 uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */ 268 uint32_t t_end_info_status; /* Status flag of end info */ 269 unsigned int *t_tfo_pending; /* TCP Fast Open server pending counter */ 270 union { 271 uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN]; 272 uint64_t server; 273 } t_tfo_cookie; /* TCP Fast Open cookie to send */ 274 union { 275 uint8_t t_end_info_bytes[TCP_END_BYTE_INFO]; 276 uint64_t t_end_info; 277 }; 278 #ifdef TCPPCAP 279 struct mbufq t_inpkts; /* List of saved input packets. */ 280 struct mbufq t_outpkts; /* List of saved output packets. */ 281 #endif 282 }; 283 #endif /* _KERNEL || _WANT_TCPCB */ 284 285 #ifdef _KERNEL 286 struct tcptemp { 287 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ 288 struct tcphdr tt_t; 289 }; 290 291 /* Enable TCP/UDP tunneling port */ 292 #define TCP_TUNNELING_PORT_MIN 0 293 #define TCP_TUNNELING_PORT_MAX 65535 294 #define TCP_TUNNELING_PORT_DEFAULT 0 295 296 /* Enable TCP/UDP tunneling port */ 297 #define TCP_TUNNELING_OVERHEAD_MIN sizeof(struct udphdr) 298 #define TCP_TUNNELING_OVERHEAD_MAX 1024 299 #define TCP_TUNNELING_OVERHEAD_DEFAULT TCP_TUNNELING_OVERHEAD_MIN 300 301 /* Minimum map entries limit value, if set */ 302 #define TCP_MIN_MAP_ENTRIES_LIMIT 128 303 304 /* 305 * TODO: We yet need to brave plowing in 306 * to tcp_input() and the pru_usrreq() block. 307 * Right now these go to the old standards which 308 * are somewhat ok, but in the long term may 309 * need to be changed. If we do tackle tcp_input() 310 * then we need to get rid of the tcp_do_segment() 311 * function below. 312 */ 313 /* Flags for tcp functions */ 314 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */ 315 316 /* 317 * If defining the optional tcp_timers, in the 318 * tfb_tcp_timer_stop call you must use the 319 * callout_async_drain() function with the 320 * tcp_timer_discard callback. You should check 321 * the return of callout_async_drain() and if 0 322 * increment tt_draincnt. Since the timer sub-system 323 * does not know your callbacks you must provide a 324 * stop_all function that loops through and calls 325 * tcp_timer_stop() with each of your defined timers. 326 * Adding a tfb_tcp_handoff_ok function allows the socket 327 * option to change stacks to query you even if the 328 * connection is in a later stage. You return 0 to 329 * say you can take over and run your stack, you return 330 * non-zero (an error number) to say no you can't. 331 * If the function is undefined you can only change 332 * in the early states (before connect or listen). 333 * tfb_tcp_fb_fini is changed to add a flag to tell 334 * the old stack if the tcb is being destroyed or 335 * not. A one in the flag means the TCB is being 336 * destroyed, a zero indicates its transitioning to 337 * another stack (via socket option). 338 */ 339 struct tcp_function_block { 340 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX]; 341 int (*tfb_tcp_output)(struct tcpcb *); 342 int (*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *); 343 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *, 344 struct socket *, struct tcpcb *, 345 int, int, uint8_t); 346 int (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int); 347 int (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *, 348 struct socket *, struct tcpcb *, 349 int, int, uint8_t, 350 int, struct timeval *); 351 void (*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *, 352 struct socket *, struct tcpcb *, 353 int, int, uint8_t, 354 int, struct timeval *); 355 int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt, 356 struct inpcb *inp, struct tcpcb *tp); 357 /* Optional memory allocation/free routine */ 358 int (*tfb_tcp_fb_init)(struct tcpcb *); 359 void (*tfb_tcp_fb_fini)(struct tcpcb *, int); 360 /* Optional timers, must define all if you define one */ 361 int (*tfb_tcp_timer_stop_all)(struct tcpcb *); 362 void (*tfb_tcp_timer_activate)(struct tcpcb *, 363 uint32_t, u_int); 364 int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t); 365 void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t); 366 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *); 367 int (*tfb_tcp_handoff_ok)(struct tcpcb *); 368 void (*tfb_tcp_mtu_chg)(struct tcpcb *); 369 int (*tfb_pru_options)(struct tcpcb *, int); 370 volatile uint32_t tfb_refcnt; 371 uint32_t tfb_flags; 372 uint8_t tfb_id; 373 }; 374 375 struct tcp_function { 376 TAILQ_ENTRY(tcp_function) tf_next; 377 char tf_name[TCP_FUNCTION_NAME_LEN_MAX]; 378 struct tcp_function_block *tf_fb; 379 }; 380 381 TAILQ_HEAD(tcp_funchead, tcp_function); 382 #endif /* _KERNEL */ 383 384 /* 385 * Flags and utility macros for the t_flags field. 386 */ 387 #define TF_ACKNOW 0x00000001 /* ack peer immediately */ 388 #define TF_DELACK 0x00000002 /* ack, but try to delay it */ 389 #define TF_NODELAY 0x00000004 /* don't delay packets to coalesce */ 390 #define TF_NOOPT 0x00000008 /* don't use tcp options */ 391 #define TF_SENTFIN 0x00000010 /* have sent FIN */ 392 #define TF_REQ_SCALE 0x00000020 /* have/will request window scaling */ 393 #define TF_RCVD_SCALE 0x00000040 /* other side has requested scaling */ 394 #define TF_REQ_TSTMP 0x00000080 /* have/will request timestamps */ 395 #define TF_RCVD_TSTMP 0x00000100 /* a timestamp was received in SYN */ 396 #define TF_SACK_PERMIT 0x00000200 /* other side said I could SACK */ 397 #define TF_NEEDSYN 0x00000400 /* send SYN (implicit state) */ 398 #define TF_NEEDFIN 0x00000800 /* send FIN (implicit state) */ 399 #define TF_NOPUSH 0x00001000 /* don't push */ 400 #define TF_PREVVALID 0x00002000 /* saved values for bad rxmit valid */ 401 #define TF_WAKESOR 0x00004000 /* wake up receive socket */ 402 #define TF_GPUTINPROG 0x00008000 /* Goodput measurement in progress */ 403 #define TF_MORETOCOME 0x00010000 /* More data to be appended to sock */ 404 #define TF_LQ_OVERFLOW 0x00020000 /* listen queue overflow */ 405 #define TF_LASTIDLE 0x00040000 /* connection was previously idle */ 406 #define TF_RXWIN0SENT 0x00080000 /* sent a receiver win 0 in response */ 407 #define TF_FASTRECOVERY 0x00100000 /* in NewReno Fast Recovery */ 408 #define TF_WASFRECOVERY 0x00200000 /* was in NewReno Fast Recovery */ 409 #define TF_SIGNATURE 0x00400000 /* require MD5 digests (RFC2385) */ 410 #define TF_FORCEDATA 0x00800000 /* force out a byte */ 411 #define TF_TSO 0x01000000 /* TSO enabled on this connection */ 412 #define TF_TOE 0x02000000 /* this connection is offloaded */ 413 #define TF_UNUSED0 0x04000000 /* unused */ 414 #define TF_UNUSED1 0x08000000 /* unused */ 415 #define TF_LRD 0x10000000 /* Lost Retransmission Detection */ 416 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */ 417 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */ 418 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */ 419 420 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY) 421 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY 422 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY 423 424 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY) 425 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY 426 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY 427 428 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY)) 429 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY) 430 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY) 431 432 #if defined(_KERNEL) && !defined(TCP_RFC7413) 433 #define IS_FASTOPEN(t_flags) (false) 434 #else 435 #define IS_FASTOPEN(t_flags) (t_flags & TF_FASTOPEN) 436 #endif 437 438 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una) 439 440 /* 441 * Flags for the t_oobflags field. 442 */ 443 #define TCPOOB_HAVEDATA 0x01 444 #define TCPOOB_HADDATA 0x02 445 446 /* 447 * Flags for the extended TCP flags field, t_flags2 448 */ 449 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */ 450 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */ 451 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */ 452 #define TF2_LOG_AUTO 0x00000008 /* Session is auto-logging. */ 453 #define TF2_DROP_AF_DATA 0x00000010 /* Drop after all data ack'd */ 454 #define TF2_ECN_PERMIT 0x00000020 /* connection ECN-ready */ 455 #define TF2_ECN_SND_CWR 0x00000040 /* ECN CWR in queue */ 456 #define TF2_ECN_SND_ECE 0x00000080 /* ECN ECE in queue */ 457 #define TF2_ACE_PERMIT 0x00000100 /* Accurate ECN mode */ 458 #define TF2_FBYTES_COMPLETE 0x00000400 /* We have first bytes in and out */ 459 /* 460 * Structure to hold TCP options that are only used during segment 461 * processing (in tcp_input), but not held in the tcpcb. 462 * It's basically used to reduce the number of parameters 463 * to tcp_dooptions and tcp_addoptions. 464 * The binary order of the to_flags is relevant for packing of the 465 * options in tcp_addoptions. 466 */ 467 struct tcpopt { 468 u_int32_t to_flags; /* which options are present */ 469 #define TOF_MSS 0x0001 /* maximum segment size */ 470 #define TOF_SCALE 0x0002 /* window scaling */ 471 #define TOF_SACKPERM 0x0004 /* SACK permitted */ 472 #define TOF_TS 0x0010 /* timestamp */ 473 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */ 474 #define TOF_SACK 0x0080 /* Peer sent SACK option */ 475 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */ 476 #define TOF_MAXOPT 0x0200 477 u_int32_t to_tsval; /* new timestamp */ 478 u_int32_t to_tsecr; /* reflected timestamp */ 479 u_char *to_sacks; /* pointer to the first SACK blocks */ 480 u_char *to_signature; /* pointer to the TCP-MD5 signature */ 481 u_int8_t *to_tfo_cookie; /* pointer to the TFO cookie */ 482 u_int16_t to_mss; /* maximum segment size */ 483 u_int8_t to_wscale; /* window scaling */ 484 u_int8_t to_nsacks; /* number of SACK blocks */ 485 u_int8_t to_tfo_len; /* TFO cookie length */ 486 u_int32_t to_spare; /* UTO */ 487 }; 488 489 /* 490 * Flags for tcp_dooptions. 491 */ 492 #define TO_SYN 0x01 /* parse SYN-only options */ 493 494 struct hc_metrics_lite { /* must stay in sync with hc_metrics */ 495 uint32_t rmx_mtu; /* MTU for this path */ 496 uint32_t rmx_ssthresh; /* outbound gateway buffer limit */ 497 uint32_t rmx_rtt; /* estimated round trip time */ 498 uint32_t rmx_rttvar; /* estimated rtt variance */ 499 uint32_t rmx_cwnd; /* congestion window */ 500 uint32_t rmx_sendpipe; /* outbound delay-bandwidth product */ 501 uint32_t rmx_recvpipe; /* inbound delay-bandwidth product */ 502 }; 503 504 /* 505 * Used by tcp_maxmtu() to communicate interface specific features 506 * and limits at the time of connection setup. 507 */ 508 struct tcp_ifcap { 509 int ifcap; 510 u_int tsomax; 511 u_int tsomaxsegcount; 512 u_int tsomaxsegsize; 513 }; 514 515 #ifndef _NETINET_IN_PCB_H_ 516 struct in_conninfo; 517 #endif /* _NETINET_IN_PCB_H_ */ 518 519 struct tcptw { 520 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */ 521 uint32_t t_port:16, /* UDP port number if TCPoUDP */ 522 t_unused:16; 523 tcp_seq snd_nxt; 524 tcp_seq rcv_nxt; 525 tcp_seq iss; 526 tcp_seq irs; 527 u_short last_win; /* cached window value */ 528 short tw_so_options; /* copy of so_options */ 529 struct ucred *tw_cred; /* user credentials */ 530 u_int32_t t_recent; 531 u_int32_t ts_offset; /* our timestamp offset */ 532 u_int t_starttime; 533 int tw_time; 534 TAILQ_ENTRY(tcptw) tw_2msl; 535 void *tw_pspare; /* TCP_SIGNATURE */ 536 u_int *tw_spare; /* TCP_SIGNATURE */ 537 }; 538 539 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) 540 #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb) 541 #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) 542 543 /* 544 * The smoothed round-trip time and estimated variance 545 * are stored as fixed point numbers scaled by the values below. 546 * For convenience, these scales are also used in smoothing the average 547 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). 548 * With these scales, srtt has 3 bits to the right of the binary point, 549 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the 550 * binary point, and is smoothed with an ALPHA of 0.75. 551 */ 552 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ 553 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ 554 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ 555 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ 556 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ 557 558 /* 559 * The initial retransmission should happen at rtt + 4 * rttvar. 560 * Because of the way we do the smoothing, srtt and rttvar 561 * will each average +1/2 tick of bias. When we compute 562 * the retransmit timer, we want 1/2 tick of rounding and 563 * 1 extra tick because of +-1/2 tick uncertainty in the 564 * firing of the timer. The bias will give us exactly the 565 * 1.5 tick we need. But, because the bias is 566 * statistical, we have to test that we don't drop below 567 * the minimum feasible timer (which is 2 ticks). 568 * This version of the macro adapted from a paper by Lawrence 569 * Brakmo and Larry Peterson which outlines a problem caused 570 * by insufficient precision in the original implementation, 571 * which results in inappropriately large RTO values for very 572 * fast networks. 573 */ 574 #define TCP_REXMTVAL(tp) \ 575 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ 576 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) 577 578 /* 579 * TCP statistics. 580 * Many of these should be kept per connection, 581 * but that's inconvenient at the moment. 582 */ 583 struct tcpstat { 584 uint64_t tcps_connattempt; /* connections initiated */ 585 uint64_t tcps_accepts; /* connections accepted */ 586 uint64_t tcps_connects; /* connections established */ 587 uint64_t tcps_drops; /* connections dropped */ 588 uint64_t tcps_conndrops; /* embryonic connections dropped */ 589 uint64_t tcps_minmssdrops; /* average minmss too low drops */ 590 uint64_t tcps_closed; /* conn. closed (includes drops) */ 591 uint64_t tcps_segstimed; /* segs where we tried to get rtt */ 592 uint64_t tcps_rttupdated; /* times we succeeded */ 593 uint64_t tcps_delack; /* delayed acks sent */ 594 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ 595 uint64_t tcps_rexmttimeo; /* retransmit timeouts */ 596 uint64_t tcps_persisttimeo; /* persist timeouts */ 597 uint64_t tcps_keeptimeo; /* keepalive timeouts */ 598 uint64_t tcps_keepprobe; /* keepalive probes sent */ 599 uint64_t tcps_keepdrops; /* connections dropped in keepalive */ 600 601 uint64_t tcps_sndtotal; /* total packets sent */ 602 uint64_t tcps_sndpack; /* data packets sent */ 603 uint64_t tcps_sndbyte; /* data bytes sent */ 604 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */ 605 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */ 606 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */ 607 uint64_t tcps_sndacks; /* ack-only packets sent */ 608 uint64_t tcps_sndprobe; /* window probes sent */ 609 uint64_t tcps_sndurg; /* packets sent with URG only */ 610 uint64_t tcps_sndwinup; /* window update-only packets sent */ 611 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ 612 613 uint64_t tcps_rcvtotal; /* total packets received */ 614 uint64_t tcps_rcvpack; /* packets received in sequence */ 615 uint64_t tcps_rcvbyte; /* bytes received in sequence */ 616 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */ 617 uint64_t tcps_rcvbadoff; /* packets received with bad offset */ 618 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */ 619 uint64_t tcps_rcvshort; /* packets received too short */ 620 uint64_t tcps_rcvduppack; /* duplicate-only packets received */ 621 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */ 622 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */ 623 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ 624 uint64_t tcps_rcvoopack; /* out-of-order packets received */ 625 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */ 626 uint64_t tcps_rcvpackafterwin; /* packets with data after window */ 627 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */ 628 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */ 629 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */ 630 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */ 631 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */ 632 uint64_t tcps_rcvackpack; /* rcvd ack packets */ 633 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */ 634 uint64_t tcps_rcvwinupd; /* rcvd window update packets */ 635 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */ 636 uint64_t tcps_predack; /* times hdr predict ok for acks */ 637 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */ 638 uint64_t tcps_pcbcachemiss; 639 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */ 640 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */ 641 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */ 642 uint64_t tcps_usedrtt; /* times RTT initialized from route */ 643 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */ 644 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/ 645 uint64_t tcps_persistdrop; /* timeout in persist state */ 646 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */ 647 uint64_t tcps_mturesent; /* resends due to MTU discovery */ 648 uint64_t tcps_listendrop; /* listen queue overflows */ 649 uint64_t tcps_badrst; /* ignored RSTs in the window */ 650 651 uint64_t tcps_sc_added; /* entry added to syncache */ 652 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */ 653 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */ 654 uint64_t tcps_sc_dropped; /* could not reply to packet */ 655 uint64_t tcps_sc_completed; /* successful extraction of entry */ 656 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */ 657 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */ 658 uint64_t tcps_sc_reset; /* RST removed entry from syncache */ 659 uint64_t tcps_sc_stale; /* timed out or listen socket gone */ 660 uint64_t tcps_sc_aborted; /* syncache entry aborted */ 661 uint64_t tcps_sc_badack; /* removed due to bad ACK */ 662 uint64_t tcps_sc_unreach; /* ICMP unreachable received */ 663 uint64_t tcps_sc_zonefail; /* zalloc() failed */ 664 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */ 665 uint64_t tcps_sc_recvcookie; /* SYN cookie received */ 666 667 uint64_t tcps_hc_added; /* entry added to hostcache */ 668 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */ 669 670 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ 671 672 /* SACK related stats */ 673 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */ 674 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */ 675 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ 676 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ 677 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ 678 uint64_t tcps_sack_lostrexmt; /* SACK lost retransmission recovered */ 679 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */ 680 681 /* ECN related stats */ 682 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */ 683 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */ 684 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */ 685 uint64_t tcps_ecn_shs; /* ECN successful handshakes */ 686 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */ 687 688 /* TCP_SIGNATURE related stats */ 689 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */ 690 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */ 691 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */ 692 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */ 693 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */ 694 695 /* Path MTU Discovery Black Hole Detection related stats */ 696 uint64_t tcps_pmtud_blackhole_activated; /* Black Hole Count */ 697 uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */ 698 uint64_t tcps_pmtud_blackhole_failed; /* Black Hole Failure Count */ 699 700 uint64_t tcps_tunneled_pkts; /* Packets encap's in UDP received */ 701 uint64_t tcps_tunneled_errs; /* Packets that had errors that were UDP encaped */ 702 703 uint64_t _pad[9]; /* 6 UTO, 3 TBD */ 704 }; 705 706 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */ 707 708 #ifdef _KERNEL 709 #define TI_UNLOCKED 1 710 #define TI_RLOCKED 2 711 #include <sys/counter.h> 712 713 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */ 714 /* 715 * In-kernel consumers can use these accessor macros directly to update 716 * stats. 717 */ 718 #define TCPSTAT_ADD(name, val) \ 719 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)) 720 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1) 721 722 /* 723 * Kernel module consumers must use this accessor macro. 724 */ 725 void kmod_tcpstat_add(int statnum, int val); 726 #define KMOD_TCPSTAT_ADD(name, val) \ 727 kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val) 728 #define KMOD_TCPSTAT_INC(name) KMOD_TCPSTAT_ADD(name, 1) 729 730 /* 731 * Running TCP connection count by state. 732 */ 733 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]); 734 #define V_tcps_states VNET(tcps_states) 735 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1) 736 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1) 737 738 /* 739 * TCP specific helper hook point identifiers. 740 */ 741 #define HHOOK_TCP_EST_IN 0 742 #define HHOOK_TCP_EST_OUT 1 743 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT 744 745 struct tcp_hhook_data { 746 struct tcpcb *tp; 747 struct tcphdr *th; 748 struct tcpopt *to; 749 uint32_t len; 750 int tso; 751 tcp_seq curack; 752 }; 753 #ifdef TCP_HHOOK 754 void hhook_run_tcp_est_out(struct tcpcb *tp, 755 struct tcphdr *th, struct tcpopt *to, 756 uint32_t len, int tso); 757 #endif 758 #endif 759 760 /* 761 * TCB structure exported to user-land via sysctl(3). 762 * 763 * Fields prefixed with "xt_" are unique to the export structure, and fields 764 * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'. 765 * 766 * Legend: 767 * (s) - used by userland utilities in src 768 * (p) - used by utilities in ports 769 * (3) - is known to be used by third party software not in ports 770 * (n) - no known usage 771 * 772 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been 773 * included. Not all of our clients do. 774 */ 775 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) 776 struct xtcpcb { 777 ksize_t xt_len; /* length of this structure */ 778 struct xinpcb xt_inp; 779 char xt_stack[TCP_FUNCTION_NAME_LEN_MAX]; /* (s) */ 780 char xt_logid[TCP_LOG_ID_LEN]; /* (s) */ 781 char xt_cc[TCP_CA_NAME_MAX]; /* (s) */ 782 int64_t spare64[6]; 783 int32_t t_state; /* (s,p) */ 784 uint32_t t_flags; /* (s,p) */ 785 int32_t t_sndzerowin; /* (s) */ 786 int32_t t_sndrexmitpack; /* (s) */ 787 int32_t t_rcvoopack; /* (s) */ 788 int32_t t_rcvtime; /* (s) */ 789 int32_t tt_rexmt; /* (s) */ 790 int32_t tt_persist; /* (s) */ 791 int32_t tt_keep; /* (s) */ 792 int32_t tt_2msl; /* (s) */ 793 int32_t tt_delack; /* (s) */ 794 int32_t t_logstate; /* (3) */ 795 uint32_t t_snd_cwnd; /* (s) */ 796 uint32_t t_snd_ssthresh; /* (s) */ 797 uint32_t t_maxseg; /* (s) */ 798 uint32_t t_rcv_wnd; /* (s) */ 799 uint32_t t_snd_wnd; /* (s) */ 800 uint32_t xt_ecn; /* (s) */ 801 uint16_t xt_encaps_port; /* (s) */ 802 int16_t spare16; 803 int32_t spare32[25]; 804 } __aligned(8); 805 806 #ifdef _KERNEL 807 void tcp_inptoxtp(const struct inpcb *, struct xtcpcb *); 808 #endif 809 #endif 810 811 /* 812 * TCP function information (name-to-id mapping, aliases, and refcnt) 813 * exported to user-land via sysctl(3). 814 */ 815 struct tcp_function_info { 816 uint32_t tfi_refcnt; 817 uint8_t tfi_id; 818 char tfi_name[TCP_FUNCTION_NAME_LEN_MAX]; 819 char tfi_alias[TCP_FUNCTION_NAME_LEN_MAX]; 820 }; 821 822 /* 823 * Identifiers for TCP sysctl nodes 824 */ 825 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ 826 #define TCPCTL_MSSDFLT 3 /* MSS default */ 827 #define TCPCTL_STATS 4 /* statistics */ 828 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */ 829 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ 830 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ 831 #define TCPCTL_SENDSPACE 8 /* send buffer space */ 832 #define TCPCTL_RECVSPACE 9 /* receive buffer space */ 833 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ 834 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ 835 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ 836 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ 837 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ 838 #define TCPCTL_DROP 15 /* drop tcp connection */ 839 #define TCPCTL_STATES 16 /* connection counts by TCP state */ 840 841 #ifdef _KERNEL 842 #ifdef SYSCTL_DECL 843 SYSCTL_DECL(_net_inet_tcp); 844 SYSCTL_DECL(_net_inet_tcp_sack); 845 MALLOC_DECLARE(M_TCPLOG); 846 #endif 847 848 VNET_DECLARE(int, tcp_log_in_vain); 849 #define V_tcp_log_in_vain VNET(tcp_log_in_vain) 850 851 /* 852 * Global TCP tunables shared between different stacks. 853 * Please keep the list sorted. 854 */ 855 VNET_DECLARE(int, drop_synfin); 856 VNET_DECLARE(int, path_mtu_discovery); 857 VNET_DECLARE(int, tcp_abc_l_var); 858 VNET_DECLARE(int, tcp_autorcvbuf_max); 859 VNET_DECLARE(int, tcp_autosndbuf_inc); 860 VNET_DECLARE(int, tcp_autosndbuf_max); 861 VNET_DECLARE(int, tcp_delack_enabled); 862 VNET_DECLARE(int, tcp_do_autorcvbuf); 863 VNET_DECLARE(int, tcp_do_autosndbuf); 864 VNET_DECLARE(int, tcp_do_ecn); 865 VNET_DECLARE(int, tcp_do_lrd); 866 VNET_DECLARE(int, tcp_do_prr); 867 VNET_DECLARE(int, tcp_do_prr_conservative); 868 VNET_DECLARE(int, tcp_do_newcwv); 869 VNET_DECLARE(int, tcp_do_rfc1323); 870 VNET_DECLARE(int, tcp_tolerate_missing_ts); 871 VNET_DECLARE(int, tcp_do_rfc3042); 872 VNET_DECLARE(int, tcp_do_rfc3390); 873 VNET_DECLARE(int, tcp_do_rfc3465); 874 VNET_DECLARE(int, tcp_do_newsack); 875 VNET_DECLARE(int, tcp_do_sack); 876 VNET_DECLARE(int, tcp_do_tso); 877 VNET_DECLARE(int, tcp_ecn_maxretries); 878 VNET_DECLARE(int, tcp_initcwnd_segments); 879 VNET_DECLARE(int, tcp_insecure_rst); 880 VNET_DECLARE(int, tcp_insecure_syn); 881 VNET_DECLARE(uint32_t, tcp_map_entries_limit); 882 VNET_DECLARE(uint32_t, tcp_map_split_limit); 883 VNET_DECLARE(int, tcp_minmss); 884 VNET_DECLARE(int, tcp_mssdflt); 885 #ifdef STATS 886 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl); 887 VNET_DECLARE(int, tcp_perconn_stats_enable); 888 #endif /* STATS */ 889 VNET_DECLARE(int, tcp_recvspace); 890 VNET_DECLARE(int, tcp_sack_globalholes); 891 VNET_DECLARE(int, tcp_sack_globalmaxholes); 892 VNET_DECLARE(int, tcp_sack_maxholes); 893 VNET_DECLARE(int, tcp_sc_rst_sock_fail); 894 VNET_DECLARE(int, tcp_sendspace); 895 VNET_DECLARE(int, tcp_udp_tunneling_overhead); 896 VNET_DECLARE(int, tcp_udp_tunneling_port); 897 VNET_DECLARE(struct inpcbhead, tcb); 898 VNET_DECLARE(struct inpcbinfo, tcbinfo); 899 900 #define V_tcp_do_lrd VNET(tcp_do_lrd) 901 #define V_tcp_do_prr VNET(tcp_do_prr) 902 #define V_tcp_do_prr_conservative VNET(tcp_do_prr_conservative) 903 #define V_tcp_do_newcwv VNET(tcp_do_newcwv) 904 #define V_drop_synfin VNET(drop_synfin) 905 #define V_path_mtu_discovery VNET(path_mtu_discovery) 906 #define V_tcb VNET(tcb) 907 #define V_tcbinfo VNET(tcbinfo) 908 #define V_tcp_abc_l_var VNET(tcp_abc_l_var) 909 #define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max) 910 #define V_tcp_autosndbuf_inc VNET(tcp_autosndbuf_inc) 911 #define V_tcp_autosndbuf_max VNET(tcp_autosndbuf_max) 912 #define V_tcp_delack_enabled VNET(tcp_delack_enabled) 913 #define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf) 914 #define V_tcp_do_autosndbuf VNET(tcp_do_autosndbuf) 915 #define V_tcp_do_ecn VNET(tcp_do_ecn) 916 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323) 917 #define V_tcp_tolerate_missing_ts VNET(tcp_tolerate_missing_ts) 918 #define V_tcp_ts_offset_per_conn VNET(tcp_ts_offset_per_conn) 919 #define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042) 920 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390) 921 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465) 922 #define V_tcp_do_newsack VNET(tcp_do_newsack) 923 #define V_tcp_do_sack VNET(tcp_do_sack) 924 #define V_tcp_do_tso VNET(tcp_do_tso) 925 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries) 926 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments) 927 #define V_tcp_insecure_rst VNET(tcp_insecure_rst) 928 #define V_tcp_insecure_syn VNET(tcp_insecure_syn) 929 #define V_tcp_map_entries_limit VNET(tcp_map_entries_limit) 930 #define V_tcp_map_split_limit VNET(tcp_map_split_limit) 931 #define V_tcp_minmss VNET(tcp_minmss) 932 #define V_tcp_mssdflt VNET(tcp_mssdflt) 933 #ifdef STATS 934 #define V_tcp_perconn_stats_dflt_tpl VNET(tcp_perconn_stats_dflt_tpl) 935 #define V_tcp_perconn_stats_enable VNET(tcp_perconn_stats_enable) 936 #endif /* STATS */ 937 #define V_tcp_recvspace VNET(tcp_recvspace) 938 #define V_tcp_sack_globalholes VNET(tcp_sack_globalholes) 939 #define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes) 940 #define V_tcp_sack_maxholes VNET(tcp_sack_maxholes) 941 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail) 942 #define V_tcp_sendspace VNET(tcp_sendspace) 943 #define V_tcp_udp_tunneling_overhead VNET(tcp_udp_tunneling_overhead) 944 #define V_tcp_udp_tunneling_port VNET(tcp_udp_tunneling_port) 945 946 #ifdef TCP_HHOOK 947 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]); 948 #define V_tcp_hhh VNET(tcp_hhh) 949 #endif 950 951 int tcp_addoptions(struct tcpopt *, u_char *); 952 int tcp_ccalgounload(struct cc_algo *unload_algo); 953 struct tcpcb * 954 tcp_close(struct tcpcb *); 955 void tcp_discardcb(struct tcpcb *); 956 void tcp_twstart(struct tcpcb *); 957 void tcp_twclose(struct tcptw *, int); 958 void tcp_ctlinput(int, struct sockaddr *, void *); 959 int tcp_ctloutput(struct socket *, struct sockopt *); 960 void tcp_ctlinput_viaudp(int, struct sockaddr *, void *, void *); 961 struct tcpcb * 962 tcp_drop(struct tcpcb *, int); 963 void tcp_drain(void); 964 void tcp_init(void); 965 void tcp_fini(void *); 966 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *, 967 const void *); 968 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *, 969 const void *); 970 int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *, 971 struct mbuf *); 972 void tcp_reass_global_init(void); 973 void tcp_reass_flush(struct tcpcb *); 974 void tcp_dooptions(struct tcpopt *, u_char *, int, int); 975 void tcp_dropwithreset(struct mbuf *, struct tcphdr *, 976 struct tcpcb *, int, int); 977 void tcp_pulloutofband(struct socket *, 978 struct tcphdr *, struct mbuf *, int); 979 void tcp_xmit_timer(struct tcpcb *, int); 980 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); 981 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th, 982 uint16_t nsegs, uint16_t type); 983 void cc_conn_init(struct tcpcb *tp); 984 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); 985 void cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos); 986 void cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos); 987 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type); 988 #ifdef TCP_HHOOK 989 void hhook_run_tcp_est_in(struct tcpcb *tp, 990 struct tcphdr *th, struct tcpopt *to); 991 #endif 992 993 int tcp_input(struct mbuf **, int *, int); 994 int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *, 995 struct tcpcb *, int); 996 int tcp_input_with_port(struct mbuf **, int *, int, uint16_t); 997 void tcp_handle_wakeup(struct tcpcb *, struct socket *); 998 void tcp_do_segment(struct mbuf *, struct tcphdr *, 999 struct socket *, struct tcpcb *, int, int, uint8_t); 1000 1001 int register_tcp_functions(struct tcp_function_block *blk, int wait); 1002 int register_tcp_functions_as_names(struct tcp_function_block *blk, 1003 int wait, const char *names[], int *num_names); 1004 int register_tcp_functions_as_name(struct tcp_function_block *blk, 1005 const char *name, int wait); 1006 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce, 1007 bool force); 1008 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs); 1009 void tcp_switch_back_to_default(struct tcpcb *tp); 1010 struct tcp_function_block * 1011 find_and_ref_tcp_fb(struct tcp_function_block *fs); 1012 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp); 1013 1014 extern counter_u64_t tcp_inp_lro_direct_queue; 1015 extern counter_u64_t tcp_inp_lro_wokeup_queue; 1016 extern counter_u64_t tcp_inp_lro_compressed; 1017 extern counter_u64_t tcp_inp_lro_locks_taken; 1018 extern counter_u64_t tcp_extra_mbuf; 1019 extern counter_u64_t tcp_would_have_but; 1020 extern counter_u64_t tcp_comp_total; 1021 extern counter_u64_t tcp_uncomp_total; 1022 1023 #ifdef NETFLIX_EXP_DETECTION 1024 /* Various SACK attack thresholds */ 1025 extern int32_t tcp_force_detection; 1026 extern int32_t tcp_sack_to_ack_thresh; 1027 extern int32_t tcp_sack_to_move_thresh; 1028 extern int32_t tcp_restoral_thresh; 1029 extern int32_t tcp_sad_decay_val; 1030 extern int32_t tcp_sad_pacing_interval; 1031 extern int32_t tcp_sad_low_pps; 1032 extern int32_t tcp_map_minimum; 1033 extern int32_t tcp_attack_on_turns_on_logging; 1034 #endif 1035 extern uint32_t tcp_ack_war_time_window; 1036 extern uint32_t tcp_ack_war_cnt; 1037 1038 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *); 1039 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *); 1040 u_int tcp_maxseg(const struct tcpcb *); 1041 u_int tcp_fixed_maxseg(const struct tcpcb *); 1042 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *, 1043 struct tcp_ifcap *); 1044 void tcp_mss(struct tcpcb *, int); 1045 int tcp_mssopt(struct in_conninfo *); 1046 struct inpcb * 1047 tcp_drop_syn_sent(struct inpcb *, int); 1048 struct tcpcb * 1049 tcp_newtcpcb(struct inpcb *); 1050 int tcp_output(struct tcpcb *); 1051 void tcp_state_change(struct tcpcb *, int); 1052 void tcp_respond(struct tcpcb *, void *, 1053 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int); 1054 void tcp_tw_init(void); 1055 #ifdef VIMAGE 1056 void tcp_tw_destroy(void); 1057 #endif 1058 void tcp_tw_zone_change(void); 1059 int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *, 1060 struct mbuf *, int); 1061 void tcp_setpersist(struct tcpcb *); 1062 void tcp_slowtimo(void); 1063 struct tcptemp * 1064 tcpip_maketemplate(struct inpcb *); 1065 void tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *); 1066 void tcp_timer_activate(struct tcpcb *, uint32_t, u_int); 1067 int tcp_timer_suspend(struct tcpcb *, uint32_t); 1068 void tcp_timers_unsuspend(struct tcpcb *, uint32_t); 1069 int tcp_timer_active(struct tcpcb *, uint32_t); 1070 void tcp_timer_stop(struct tcpcb *, uint32_t); 1071 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int); 1072 int inp_to_cpuid(struct inpcb *inp); 1073 /* 1074 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) 1075 */ 1076 void tcp_hc_init(void); 1077 #ifdef VIMAGE 1078 void tcp_hc_destroy(void); 1079 #endif 1080 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); 1081 uint32_t tcp_hc_getmtu(struct in_conninfo *); 1082 void tcp_hc_updatemtu(struct in_conninfo *, uint32_t); 1083 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); 1084 1085 extern struct pr_usrreqs tcp_usrreqs; 1086 1087 uint32_t tcp_new_ts_offset(struct in_conninfo *); 1088 tcp_seq tcp_new_isn(struct in_conninfo *); 1089 1090 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); 1091 int tcp_dsack_block_exists(struct tcpcb *); 1092 void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq); 1093 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); 1094 void tcp_clean_dsack_blocks(struct tcpcb *tp); 1095 void tcp_clean_sackreport(struct tcpcb *tp); 1096 void tcp_sack_adjust(struct tcpcb *tp); 1097 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); 1098 void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *); 1099 void tcp_lost_retransmission(struct tcpcb *, struct tcphdr *); 1100 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); 1101 void tcp_free_sackholes(struct tcpcb *tp); 1102 void tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *); 1103 int tcp_newreno(struct tcpcb *, struct tcphdr *); 1104 int tcp_compute_pipe(struct tcpcb *); 1105 uint32_t tcp_compute_initwnd(uint32_t); 1106 void tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t); 1107 int tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes, 1108 size_t seed_len); 1109 int tcp_can_enable_pacing(void); 1110 void tcp_decrement_paced_conn(void); 1111 1112 struct mbuf * 1113 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen, 1114 int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls); 1115 1116 int tcp_stats_init(void); 1117 void tcp_log_end_status(struct tcpcb *tp, uint8_t status); 1118 1119 static inline void 1120 tcp_fields_to_host(struct tcphdr *th) 1121 { 1122 1123 th->th_seq = ntohl(th->th_seq); 1124 th->th_ack = ntohl(th->th_ack); 1125 th->th_win = ntohs(th->th_win); 1126 th->th_urp = ntohs(th->th_urp); 1127 } 1128 1129 static inline void 1130 tcp_fields_to_net(struct tcphdr *th) 1131 { 1132 1133 th->th_seq = htonl(th->th_seq); 1134 th->th_ack = htonl(th->th_ack); 1135 th->th_win = htons(th->th_win); 1136 th->th_urp = htons(th->th_urp); 1137 } 1138 1139 static inline void 1140 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt, uint8_t is_tlp) 1141 { 1142 if (is_tlp) { 1143 tp->t_sndtlppack++; 1144 tp->t_sndtlpbyte += len; 1145 } 1146 /* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */ 1147 if (is_rxt) 1148 tp->t_snd_rxt_bytes += len; 1149 else 1150 tp->t_sndbytes += len; 1151 1152 } 1153 1154 #endif /* _KERNEL */ 1155 1156 #endif /* _NETINET_TCP_VAR_H_ */ 1157