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 "opt_kern_tls.h" 43 #include <net/vnet.h> 44 #include <sys/mbuf.h> 45 #include <sys/ktls.h> 46 #endif 47 48 #define TCP_END_BYTE_INFO 8 /* Bytes that makeup the "end information array" */ 49 /* Types of ending byte info */ 50 #define TCP_EI_EMPTY_SLOT 0 51 #define TCP_EI_STATUS_CLIENT_FIN 0x1 52 #define TCP_EI_STATUS_CLIENT_RST 0x2 53 #define TCP_EI_STATUS_SERVER_FIN 0x3 54 #define TCP_EI_STATUS_SERVER_RST 0x4 55 #define TCP_EI_STATUS_RETRAN 0x5 56 #define TCP_EI_STATUS_PROGRESS 0x6 57 #define TCP_EI_STATUS_PERSIST_MAX 0x7 58 #define TCP_EI_STATUS_KEEP_MAX 0x8 59 #define TCP_EI_STATUS_DATA_A_CLOSE 0x9 60 #define TCP_EI_STATUS_RST_IN_FRONT 0xa 61 #define TCP_EI_STATUS_2MSL 0xb 62 #define TCP_EI_STATUS_MAX_VALUE 0xb 63 64 #define TCP_HTTP_REQ_LOG_NEW 0x01 65 #define TCP_HTTP_REQ_LOG_COMPLETE 0x02 66 #define TCP_HTTP_REQ_LOG_FREED 0x03 67 #define TCP_HTTP_REQ_LOG_ALLOCFAIL 0x04 68 #define TCP_HTTP_REQ_LOG_MOREYET 0x05 69 #define TCP_HTTP_REQ_LOG_FORCEFREE 0x06 70 #define TCP_HTTP_REQ_LOG_STALE 0x07 71 #define TCP_HTTP_REQ_LOG_SEARCH 0x08 72 73 /************************************************/ 74 /* Status bits we track to assure no duplicates, 75 * the bits here are not used by the code but 76 * for human representation. To check a bit we 77 * take and shift over by 1 minus the value (1-8). 78 */ 79 /************************************************/ 80 #define TCP_EI_BITS_CLIENT_FIN 0x001 81 #define TCP_EI_BITS_CLIENT_RST 0x002 82 #define TCP_EI_BITS_SERVER_FIN 0x004 83 #define TCP_EI_BITS_SERVER_RST 0x008 84 #define TCP_EI_BITS_RETRAN 0x010 85 #define TCP_EI_BITS_PROGRESS 0x020 86 #define TCP_EI_BITS_PRESIST_MAX 0x040 87 #define TCP_EI_BITS_KEEP_MAX 0x080 88 #define TCP_EI_BITS_DATA_A_CLO 0x100 89 #define TCP_EI_BITS_RST_IN_FR 0x200 /* a front state reset */ 90 #define TCP_EI_BITS_2MS_TIMER 0x400 /* 2 MSL timer expired */ 91 92 #if defined(_KERNEL) || defined(_WANT_TCPCB) 93 #include <netinet/cc/cc.h> 94 95 /* TCP segment queue entry */ 96 struct tseg_qent { 97 TAILQ_ENTRY(tseg_qent) tqe_q; 98 struct mbuf *tqe_m; /* mbuf contains packet */ 99 struct mbuf *tqe_last; /* last mbuf in chain */ 100 tcp_seq tqe_start; /* TCP Sequence number start */ 101 int tqe_len; /* TCP segment data length */ 102 uint32_t tqe_flags; /* The flags from tcp_get_flags() */ 103 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */ 104 }; 105 TAILQ_HEAD(tsegqe_head, tseg_qent); 106 107 struct sackblk { 108 tcp_seq start; /* start seq no. of sack block */ 109 tcp_seq end; /* end seq no. */ 110 }; 111 112 struct sackhole { 113 tcp_seq start; /* start seq no. of hole */ 114 tcp_seq end; /* end seq no. */ 115 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ 116 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ 117 }; 118 119 struct sackhint { 120 struct sackhole *nexthole; 121 int32_t sack_bytes_rexmit; 122 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */ 123 124 int32_t delivered_data; /* Newly acked data from last SACK */ 125 126 int32_t sacked_bytes; /* Total sacked bytes reported by the 127 * receiver via sack option 128 */ 129 uint32_t recover_fs; /* Flight Size at the start of Loss recovery */ 130 uint32_t prr_delivered; /* Total bytes delivered using PRR */ 131 uint32_t prr_out; /* Bytes sent during IN_RECOVERY */ 132 }; 133 134 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq) 135 136 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem); 137 138 #define TCP_HTTP_TRACK_FLG_EMPTY 0x00 /* Available */ 139 #define TCP_HTTP_TRACK_FLG_USED 0x01 /* In use */ 140 #define TCP_HTTP_TRACK_FLG_OPEN 0x02 /* End is not valid (open range request) */ 141 #define TCP_HTTP_TRACK_FLG_SEQV 0x04 /* We had a sendfile that touched it */ 142 #define TCP_HTTP_TRACK_FLG_COMP 0x08 /* Sendfile as placed the last bits (range req only) */ 143 #define TCP_HTTP_TRACK_FLG_FSND 0x10 /* First send has been done into the seq space */ 144 #define MAX_TCP_HTTP_REQ 5 /* Max we will have at once */ 145 146 #ifdef TCP_REQUEST_TRK 147 struct http_sendfile_track { 148 uint64_t timestamp; /* User sent timestamp */ 149 uint64_t start; /* Start of sendfile offset */ 150 uint64_t end; /* End if not open-range req */ 151 uint64_t localtime; /* Time we actually got the req */ 152 uint64_t deadline; /* If in CU mode, deadline to delivery */ 153 uint64_t first_send; /* Time of first send in the range */ 154 uint64_t cspr; /* Client suggested pace rate */ 155 uint64_t sent_at_fs; /* What was t_sndbytes as we begun sending */ 156 uint64_t rxt_at_fs; /* What was t_snd_rxt_bytes as we begun sending */ 157 tcp_seq start_seq; /* First TCP Seq assigned */ 158 tcp_seq end_seq; /* If range req last seq */ 159 uint32_t flags; /* Type of request open etc */ 160 uint32_t sbcc_at_s; /* When we allocate what is the sb_cc */ 161 uint32_t hint_maxseg; /* Client hinted maxseg */ 162 uint32_t hybrid_flags; /* Hybrid flags on this request */ 163 }; 164 165 #endif 166 167 /* 168 * Change Query responses for a stack switch we create a structure 169 * that allows query response from the new stack to the old, if 170 * supported. 171 * 172 * There are three queries currently defined. 173 * - sendmap 174 * - timers 175 * - rack_times 176 * 177 * For the sendmap query the caller fills in the 178 * req and the req_param as the first seq (usually 179 * snd_una). When the response comes back indicating 180 * that there was data (return value 1), then the caller 181 * can build a sendmap entry based on the range and the 182 * times. The next query would then be done at the 183 * newly created sendmap_end. Repeated until sendmap_end == snd_max. 184 * 185 * Flags in sendmap_flags are defined below as well. 186 * 187 * For timers the standard PACE_TMR_XXXX flags are returned indicating 188 * a pacing timer (possibly) and one other timer. If pacing timer then 189 * the expiration timeout time in microseconds is in timer_pacing_to. 190 * And the value used with whatever timer (if a flag is set) is in 191 * timer_rxt. If no timers are running a 0 is returned and of 192 * course no flags are set in timer_hpts_flags. 193 * 194 * The rack_times are a misc collection of information that 195 * the old stack might possibly fill in. Of course its possible 196 * that an old stack may not have a piece of information. If so 197 * then setting that value to zero is advised. Setting any 198 * timestamp passed should only place a zero in it when it 199 * is unfilled. This may mean that a time is off by a micro-second 200 * but this is ok in the grand scheme of things. 201 * 202 * When switching stacks it is desireable to get as much information 203 * from the old stack to the new stack as possible. Though not always 204 * will the stack be compatible in the types of information. The 205 * init() function needs to take care when it begins changing 206 * things such as inp_flags2 and the timer units to position these 207 * changes at a point where it is unlikely they will fail after 208 * making such changes. A stack optionally can have an "undo" 209 * function 210 * 211 * To transfer information to the old stack from the new in 212 * respect to LRO and the inp_flags2, the new stack should set 213 * the inp_flags2 to what it supports. The old stack in its 214 * fini() function should call the tcp_handle_orphaned_packets() 215 * to clean up any packets. Note that a new stack should attempt 216 */ 217 218 /* Query types */ 219 #define TCP_QUERY_SENDMAP 1 220 #define TCP_QUERY_TIMERS_UP 2 221 #define TCP_QUERY_RACK_TIMES 3 222 223 /* Flags returned in sendmap_flags */ 224 #define SNDMAP_ACKED 0x000001/* The remote endpoint acked this */ 225 #define SNDMAP_OVERMAX 0x000008/* We have more retran's then we can fit */ 226 #define SNDMAP_SACK_PASSED 0x000010/* A sack was done above this block */ 227 #define SNDMAP_HAS_FIN 0x000040/* segment is sent with fin */ 228 #define SNDMAP_TLP 0x000080/* segment sent as tail-loss-probe */ 229 #define SNDMAP_HAS_SYN 0x000800/* SYN is on this guy */ 230 #define SNDMAP_HAD_PUSH 0x008000/* Push was sent on original send */ 231 #define SNDMAP_MASK (SNDMAP_ACKED|SNDMAP_OVERMAX|SNDMAP_SACK_PASSED|SNDMAP_HAS_FIN\ 232 |SNDMAP_TLP|SNDMAP_HAS_SYN|SNDMAP_HAD_PUSH) 233 #define SNDMAP_NRTX 3 234 235 struct tcp_query_resp { 236 int req; 237 uint32_t req_param; 238 union { 239 struct { 240 tcp_seq sendmap_start; 241 tcp_seq sendmap_end; 242 int sendmap_send_cnt; 243 uint64_t sendmap_time[SNDMAP_NRTX]; 244 uint64_t sendmap_ack_arrival; 245 int sendmap_flags; 246 uint32_t sendmap_r_rtr_bytes; 247 /* If FAS is available if not 0 */ 248 uint32_t sendmap_fas; 249 uint8_t sendmap_dupacks; 250 }; 251 struct { 252 uint32_t timer_hpts_flags; 253 uint32_t timer_pacing_to; 254 uint32_t timer_timer_exp; 255 }; 256 struct { 257 /* Timestamps and rtt's */ 258 uint32_t rack_reorder_ts; /* Last uscts that reordering was seen */ 259 uint32_t rack_num_dsacks; /* Num of dsacks seen */ 260 uint32_t rack_rxt_last_time; /* Last time a RXT/TLP or rack tmr went off */ 261 uint32_t rack_min_rtt; /* never 0 smallest rtt seen */ 262 uint32_t rack_rtt; /* Last rtt used by rack */ 263 uint32_t rack_tmit_time; /* The time the rtt seg was tmited */ 264 uint32_t rack_time_went_idle; /* If in persist the time we went idle */ 265 /* Prr data */ 266 uint32_t rack_sacked; 267 uint32_t rack_holes_rxt; 268 uint32_t rack_prr_delivered; 269 uint32_t rack_prr_recovery_fs; 270 uint32_t rack_prr_out; 271 uint32_t rack_prr_sndcnt; 272 /* TLP data */ 273 uint16_t rack_tlp_cnt_out; /* How many tlp's have been sent */ 274 /* Various bits */ 275 uint8_t rack_tlp_out; /* Is a TLP outstanding */ 276 uint8_t rack_srtt_measured; /* The previous stack has measured srtt */ 277 uint8_t rack_in_persist; /* Is the old stack in persists? */ 278 uint8_t rack_wanted_output; /* Did the prevous stack have a want output set */ 279 }; 280 }; 281 }; 282 283 #define TCP_TMR_GRANULARITY_TICKS 1 /* TCP timers are in ticks (msec if hz=1000) */ 284 #define TCP_TMR_GRANULARITY_USEC 2 /* TCP timers are in microseconds */ 285 286 typedef enum { 287 TT_REXMT = 0, 288 TT_PERSIST, 289 TT_KEEP, 290 TT_2MSL, 291 TT_DELACK, 292 TT_N, 293 } tt_which; 294 295 typedef enum { 296 TT_PROCESSING = 0, 297 TT_PROCESSED, 298 TT_STARTING, 299 TT_STOPPING, 300 } tt_what; 301 302 /* 303 * Tcp control block, one per tcp connection. 304 */ 305 struct tcpcb { 306 struct inpcb t_inpcb; /* embedded protocol independent cb */ 307 #define t_start_zero t_fb 308 #define t_zero_size (sizeof(struct tcpcb) - \ 309 offsetof(struct tcpcb, t_start_zero)) 310 struct tcp_function_block *t_fb;/* TCP function call block */ 311 void *t_fb_ptr; /* Pointer to t_fb specific data */ 312 313 struct callout t_callout; 314 sbintime_t t_timers[TT_N]; 315 sbintime_t t_precisions[TT_N]; 316 317 uint32_t t_maxseg:24, /* maximum segment size */ 318 _t_logstate:8; /* State of "black box" logging */ 319 uint32_t t_port:16, /* Tunneling (over udp) port */ 320 t_state:4, /* state of this connection */ 321 t_idle_reduce : 1, 322 t_delayed_ack: 7, /* Delayed ack variable */ 323 t_fin_is_rst: 1, /* Are fin's treated as resets */ 324 t_log_state_set: 1, 325 bits_spare : 2; 326 u_int t_flags; 327 tcp_seq snd_una; /* sent but unacknowledged */ 328 tcp_seq snd_max; /* highest sequence number sent; 329 * used to recognize retransmits 330 */ 331 tcp_seq snd_nxt; /* send next */ 332 tcp_seq snd_up; /* send urgent pointer */ 333 uint32_t snd_wnd; /* send window */ 334 uint32_t snd_cwnd; /* congestion-controlled window */ 335 uint32_t t_peakrate_thr; /* pre-calculated peak rate threshold */ 336 uint32_t ts_offset; /* our timestamp offset */ 337 uint32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ 338 int rcv_numsacks; /* # distinct sack blks present */ 339 u_int t_tsomax; /* TSO total burst length limit */ 340 u_int t_tsomaxsegcount; /* TSO maximum segment count */ 341 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */ 342 tcp_seq rcv_nxt; /* receive next */ 343 tcp_seq rcv_adv; /* advertised window */ 344 uint32_t rcv_wnd; /* receive window */ 345 u_int t_flags2; /* More tcpcb flags storage */ 346 int t_srtt; /* smoothed round-trip time */ 347 int t_rttvar; /* variance in round-trip time */ 348 uint32_t ts_recent; /* timestamp echo data */ 349 u_char snd_scale; /* window scaling for send window */ 350 u_char rcv_scale; /* window scaling for recv window */ 351 u_char snd_limited; /* segments limited transmitted */ 352 u_char request_r_scale; /* pending window scaling */ 353 tcp_seq last_ack_sent; 354 u_int t_rcvtime; /* inactivity time */ 355 tcp_seq rcv_up; /* receive urgent pointer */ 356 int t_segqlen; /* segment reassembly queue length */ 357 uint32_t t_segqmbuflen; /* total reassembly queue byte length */ 358 struct tsegqe_head t_segq; /* segment reassembly queue */ 359 struct mbuf *t_in_pkt; 360 struct mbuf *t_tail_pkt; 361 uint32_t snd_ssthresh; /* snd_cwnd size threshold for 362 * for slow start exponential to 363 * linear switch 364 */ 365 tcp_seq snd_wl1; /* window update seg seq number */ 366 tcp_seq snd_wl2; /* window update seg ack number */ 367 368 tcp_seq irs; /* initial receive sequence number */ 369 tcp_seq iss; /* initial send sequence number */ 370 u_int t_acktime; /* RACK and BBR incoming new data was acked */ 371 u_int t_sndtime; /* time last data was sent */ 372 u_int ts_recent_age; /* when last updated */ 373 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ 374 char t_oobflags; /* have some */ 375 char t_iobc; /* input character */ 376 uint8_t t_nic_ktls_xmit:1, /* active nic ktls xmit sessions */ 377 t_nic_ktls_xmit_dis:1, /* disabled nic xmit ktls? */ 378 t_nic_ktls_spare:6; /* spare nic ktls */ 379 int t_rxtcur; /* current retransmit value (ticks) */ 380 381 int t_rxtshift; /* log(2) of rexmt exp. backoff */ 382 u_int t_rtttime; /* RTT measurement start time */ 383 384 tcp_seq t_rtseq; /* sequence number being timed */ 385 u_int t_starttime; /* time connection was established */ 386 u_int t_fbyte_in; /* ticks time first byte queued in */ 387 u_int t_fbyte_out; /* ticks time first byte queued out */ 388 389 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */ 390 int t_blackhole_enter; /* when to enter blackhole detection */ 391 int t_blackhole_exit; /* when to exit blackhole detection */ 392 u_int t_rttmin; /* minimum rtt allowed */ 393 394 int t_softerror; /* possible error not yet reported */ 395 uint32_t max_sndwnd; /* largest window peer has offered */ 396 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */ 397 uint32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */ 398 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ 399 int t_sndzerowin; /* zero-window updates sent */ 400 int snd_numholes; /* number of holes seen by sender */ 401 u_int t_badrxtwin; /* window for retransmit recovery */ 402 TAILQ_HEAD(sackhole_head, sackhole) snd_holes; 403 /* SACK scoreboard (sorted) */ 404 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ 405 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ 406 struct sackhint sackhint; /* SACK scoreboard hint */ 407 int t_rttlow; /* smallest observerved RTT */ 408 int rfbuf_cnt; /* recv buffer autoscaling byte count */ 409 struct toedev *tod; /* toedev handling this connection */ 410 int t_sndrexmitpack; /* retransmit packets sent */ 411 int t_rcvoopack; /* out-of-order packets received */ 412 void *t_toe; /* TOE pcb pointer */ 413 struct cc_algo *t_cc; /* congestion control algorithm */ 414 struct cc_var t_ccv; /* congestion control specific vars */ 415 int t_bytes_acked; /* # bytes acked during current RTT */ 416 u_int t_maxunacktime; 417 u_int t_keepinit; /* time to establish connection */ 418 u_int t_keepidle; /* time before keepalive probes begin */ 419 u_int t_keepintvl; /* interval between keepalives */ 420 u_int t_keepcnt; /* number of keepalives before close */ 421 int t_dupacks; /* consecutive dup acks recd */ 422 int t_lognum; /* Number of log entries */ 423 int t_loglimit; /* Maximum number of log entries */ 424 uint32_t t_rcep; /* Number of received CE marked pkts */ 425 uint32_t t_scep; /* Synced number of delivered CE pkts */ 426 int64_t t_pacing_rate; /* bytes / sec, -1 => unlimited */ 427 struct tcp_log_stailq t_logs; /* Log buffer */ 428 struct tcp_log_id_node *t_lin; 429 struct tcp_log_id_bucket *t_lib; 430 const char *t_output_caller; /* Function that called tcp_output */ 431 struct statsblob *t_stats; /* Per-connection stats */ 432 /* Should these be a pointer to the arrays or an array? */ 433 #ifdef TCP_ACCOUNTING 434 uint64_t tcp_cnt_counters[TCP_NUM_CNT_COUNTERS]; 435 uint64_t tcp_proc_time[TCP_NUM_CNT_COUNTERS]; 436 #endif 437 #ifdef TCP_REQUEST_TRK 438 uint32_t tcp_hybrid_start; /* Num of times we started hybrid pacing */ 439 uint32_t tcp_hybrid_stop; /* Num of times we stopped hybrid pacing */ 440 uint32_t tcp_hybrid_error; /* Num of times we failed to start hybrid pacing */ 441 #endif 442 uint32_t t_logsn; /* Log "serial number" */ 443 uint32_t gput_ts; /* Time goodput measurement started */ 444 tcp_seq gput_seq; /* Outbound measurement seq */ 445 tcp_seq gput_ack; /* Inbound measurement ack */ 446 int32_t t_stats_gput_prev; /* XXXLAS: Prev gput measurement */ 447 uint32_t t_maxpeakrate; /* max peak rate set by user, bytes/s */ 448 uint32_t t_sndtlppack; /* tail loss probe packets sent */ 449 uint64_t t_sndtlpbyte; /* total tail loss probe bytes sent */ 450 uint64_t t_sndbytes; /* total bytes sent */ 451 uint64_t t_snd_rxt_bytes; /* total bytes retransmitted */ 452 uint32_t t_dsack_bytes; /* dsack bytes received */ 453 uint32_t t_dsack_tlp_bytes; /* dsack bytes received for TLPs sent */ 454 uint32_t t_dsack_pack; /* dsack packets we have eceived */ 455 uint8_t t_tmr_granularity; /* Granularity of all timers srtt etc */ 456 uint8_t t_rttupdated; /* number of times rtt sampled */ 457 /* TCP Fast Open */ 458 uint8_t t_tfo_client_cookie_len; /* TFO client cookie length */ 459 uint32_t t_end_info_status; /* Status flag of end info */ 460 unsigned int *t_tfo_pending; /* TFO server pending counter */ 461 union { 462 uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN]; 463 uint64_t server; 464 } t_tfo_cookie; /* TCP Fast Open cookie to send */ 465 union { 466 uint8_t t_end_info_bytes[TCP_END_BYTE_INFO]; 467 uint64_t t_end_info; 468 }; 469 #ifdef TCPPCAP 470 struct mbufq t_inpkts; /* List of saved input packets. */ 471 struct mbufq t_outpkts; /* List of saved output packets. */ 472 #endif 473 #ifdef TCP_HHOOK 474 struct osd t_osd; /* storage for Khelp module data */ 475 #endif 476 uint8_t _t_logpoint; /* Used when a BB log points is enabled */ 477 #ifdef TCP_REQUEST_TRK 478 /* Response tracking addons. */ 479 uint8_t t_http_req; /* Request count */ 480 uint8_t t_http_open; /* Number of open range requests */ 481 uint8_t t_http_closed; /* Number of closed range requests */ 482 struct http_sendfile_track t_http_info[MAX_TCP_HTTP_REQ]; 483 #endif 484 }; 485 #endif /* _KERNEL || _WANT_TCPCB */ 486 487 #ifdef _KERNEL 488 struct tcptemp { 489 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ 490 struct tcphdr tt_t; 491 }; 492 493 /* Enable TCP/UDP tunneling port */ 494 #define TCP_TUNNELING_PORT_MIN 0 495 #define TCP_TUNNELING_PORT_MAX 65535 496 #define TCP_TUNNELING_PORT_DEFAULT 0 497 498 /* Enable TCP/UDP tunneling port */ 499 #define TCP_TUNNELING_OVERHEAD_MIN sizeof(struct udphdr) 500 #define TCP_TUNNELING_OVERHEAD_MAX 1024 501 #define TCP_TUNNELING_OVERHEAD_DEFAULT TCP_TUNNELING_OVERHEAD_MIN 502 503 /* Minimum map entries limit value, if set */ 504 #define TCP_MIN_MAP_ENTRIES_LIMIT 128 505 506 /* 507 * TODO: We yet need to brave plowing in 508 * to tcp_input() and the pru_usrreq() block. 509 * Right now these go to the old standards which 510 * are somewhat ok, but in the long term may 511 * need to be changed. If we do tackle tcp_input() 512 * then we need to get rid of the tcp_do_segment() 513 * function below. 514 */ 515 /* Flags for tcp functions */ 516 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */ 517 #define TCP_FUNC_OUTPUT_CANDROP 0x02 /* tfb_tcp_output may ask tcp_drop */ 518 519 /** 520 * If defining the optional tcp_timers, in the 521 * tfb_tcp_timer_stop call you must use the 522 * callout_async_drain() function with the 523 * tcp_timer_discard callback. You should check 524 * the return of callout_async_drain() and if 0 525 * increment tt_draincnt. Since the timer sub-system 526 * does not know your callbacks you must provide a 527 * stop_all function that loops through and calls 528 * tcp_timer_stop() with each of your defined timers. 529 * 530 * Adding a tfb_tcp_handoff_ok function allows the socket 531 * option to change stacks to query you even if the 532 * connection is in a later stage. You return 0 to 533 * say you can take over and run your stack, you return 534 * non-zero (an error number) to say no you can't. 535 * If the function is undefined you can only change 536 * in the early states (before connect or listen). 537 * 538 * tfb_tcp_fb_init is used to allow the new stack to 539 * setup its control block. Among the things it must 540 * do is: 541 * a) Make sure that the inp_flags2 is setup correctly 542 * for LRO. There are two flags that the previous 543 * stack may have set INP_MBUF_ACKCMP and 544 * INP_SUPPORTS_MBUFQ. If the new stack does not 545 * support these it *should* clear the flags. 546 * b) Make sure that the timers are in the proper 547 * granularity that the stack wants. The stack 548 * should check the t_tmr_granularity field. Currently 549 * there are two values that it may hold 550 * TCP_TMR_GRANULARITY_TICKS and TCP_TMR_GRANULARITY_USEC. 551 * Use the functions tcp_timer_convert(tp, granularity); 552 * to move the timers to the correct format for your stack. 553 * 554 * The new stack may also optionally query the tfb_chg_query 555 * function if the old stack has one. The new stack may ask 556 * for one of three entries and can also state to the old 557 * stack its support for the INP_MBUF_ACKCMP and 558 * INP_SUPPORTS_MBUFQ. This is important since if there are 559 * queued ack's without that statement the old stack will 560 * be forced to discard the queued acks. The requests that 561 * can be made for information by the new stacks are: 562 * 563 * Note also that the tfb_tcp_fb_init() when called can 564 * determine if a query is needed by looking at the 565 * value passed in the ptr. The ptr is designed to be 566 * set in with any allocated memory, but the address 567 * of the condtion (ptr == &tp->t_fb_ptr) will be 568 * true if this is not a stack switch but the initial 569 * setup of a tcb (which means no query would be needed). 570 * If, however, the value is not t_fb_ptr, then the caller 571 * is in the middle of a stack switch and is the new stack. 572 * A query would be appropriate (if the new stack support 573 * the query mechanism). 574 * 575 * TCP_QUERY_SENDMAP - Query of outstanding data. 576 * TCP_QUERY_TIMERS_UP - Query about running timers. 577 * TCP_SUPPORTED_LRO - Declaration in req_param of 578 * the inp_flags2 supported by 579 * the new stack. 580 * TCP_QUERY_RACK_TIMES - Enquire about various timestamps 581 * and states the old stack may be in. 582 * 583 * tfb_tcp_fb_fini is changed to add a flag to tell 584 * the old stack if the tcb is being destroyed or 585 * not. A one in the flag means the TCB is being 586 * destroyed, a zero indicates its transitioning to 587 * another stack (via socket option). The 588 * tfb_tcp_fb_fini() function itself should not change timers 589 * or inp_flags2 (the tfb_tcp_fb_init() must do that). However 590 * if the old stack supports the LRO mbuf queuing, and the new 591 * stack does not communicate via chg messages that it too does, 592 * it must assume it does not and free any queued mbufs. 593 * 594 */ 595 struct tcp_function_block { 596 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX]; 597 int (*tfb_tcp_output)(struct tcpcb *); 598 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *, 599 struct socket *, struct tcpcb *, 600 int, int, uint8_t); 601 int (*tfb_do_queued_segments)(struct socket *, struct tcpcb *, int); 602 int (*tfb_do_segment_nounlock)(struct mbuf *, struct tcphdr *, 603 struct socket *, struct tcpcb *, 604 int, int, uint8_t, 605 int, struct timeval *); 606 void (*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *, 607 struct socket *, struct tcpcb *, 608 int, int, uint8_t, 609 int, struct timeval *); 610 int (*tfb_tcp_ctloutput)(struct inpcb *inp, struct sockopt *sopt); 611 /* Optional memory allocation/free routine */ 612 int (*tfb_tcp_fb_init)(struct tcpcb *, void **); 613 void (*tfb_tcp_fb_fini)(struct tcpcb *, int); 614 /* Optional timers, must define all if you define one */ 615 int (*tfb_tcp_timer_stop_all)(struct tcpcb *); 616 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *); 617 int (*tfb_tcp_handoff_ok)(struct tcpcb *); 618 void (*tfb_tcp_mtu_chg)(struct tcpcb *tp); 619 int (*tfb_pru_options)(struct tcpcb *, int); 620 void (*tfb_hwtls_change)(struct tcpcb *, int); 621 int (*tfb_chg_query)(struct tcpcb *, struct tcp_query_resp *); 622 void (*tfb_switch_failed)(struct tcpcb *); 623 bool (*tfb_early_wake_check)(struct tcpcb *); 624 int (*tfb_compute_pipe)(struct tcpcb *tp); 625 volatile uint32_t tfb_refcnt; 626 uint32_t tfb_flags; 627 uint8_t tfb_id; 628 }; 629 630 struct tcp_function { 631 TAILQ_ENTRY(tcp_function) tf_next; 632 char tf_name[TCP_FUNCTION_NAME_LEN_MAX]; 633 struct tcp_function_block *tf_fb; 634 }; 635 636 TAILQ_HEAD(tcp_funchead, tcp_function); 637 638 struct tcpcb * tcp_drop(struct tcpcb *, int); 639 640 #ifdef _NETINET_IN_PCB_H_ 641 #define intotcpcb(inp) __containerof((inp), struct tcpcb, t_inpcb) 642 #define sototcpcb(so) intotcpcb(sotoinpcb(so)) 643 #define tptoinpcb(tp) (&(tp)->t_inpcb) 644 #define tptosocket(tp) (tp)->t_inpcb.inp_socket 645 646 /* 647 * tcp_output() 648 * Handles tcp_drop request from advanced stacks and reports that inpcb is 649 * gone with negative return code. 650 * Drop in replacement for the default stack. 651 */ 652 static inline int 653 tcp_output(struct tcpcb *tp) 654 { 655 struct inpcb *inp = tptoinpcb(tp); 656 int rv; 657 658 INP_WLOCK_ASSERT(inp); 659 660 rv = tp->t_fb->tfb_tcp_output(tp); 661 if (rv < 0) { 662 KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP, 663 ("TCP stack %s requested tcp_drop(%p)", 664 tp->t_fb->tfb_tcp_block_name, tp)); 665 tp = tcp_drop(tp, -rv); 666 if (tp) 667 INP_WUNLOCK(inp); 668 } 669 670 return (rv); 671 } 672 673 static inline void 674 tcp_lro_features_off(struct inpcb *inp) 675 { 676 inp->inp_flags2 &= ~(INP_SUPPORTS_MBUFQ| 677 INP_MBUF_QUEUE_READY| 678 INP_DONT_SACK_QUEUE| 679 INP_MBUF_ACKCMP| 680 INP_MBUF_L_ACKS); 681 } 682 683 /* 684 * tcp_output_unlock() 685 * Always returns unlocked, handles drop request from advanced stacks. 686 * Always returns positive error code. 687 */ 688 static inline int 689 tcp_output_unlock(struct tcpcb *tp) 690 { 691 struct inpcb *inp = tptoinpcb(tp); 692 int rv; 693 694 INP_WLOCK_ASSERT(inp); 695 696 rv = tp->t_fb->tfb_tcp_output(tp); 697 if (rv < 0) { 698 KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP, 699 ("TCP stack %s requested tcp_drop(%p)", 700 tp->t_fb->tfb_tcp_block_name, tp)); 701 rv = -rv; 702 tp = tcp_drop(tp, rv); 703 if (tp) 704 INP_WUNLOCK(inp); 705 } else 706 INP_WUNLOCK(inp); 707 708 return (rv); 709 } 710 711 /* 712 * tcp_output_nodrop() 713 * Always returns locked. It is caller's responsibility to run tcp_drop()! 714 * Useful in syscall implementations, when we want to perform some logging 715 * and/or tracing with tcpcb before calling tcp_drop(). To be used with 716 * tcp_unlock_or_drop() later. 717 * 718 * XXXGL: maybe don't allow stacks to return a drop request at certain 719 * TCP states? Why would it do in connect(2)? In recv(2)? 720 */ 721 static inline int 722 tcp_output_nodrop(struct tcpcb *tp) 723 { 724 int rv; 725 726 INP_WLOCK_ASSERT(tptoinpcb(tp)); 727 728 rv = tp->t_fb->tfb_tcp_output(tp); 729 KASSERT(rv >= 0 || tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP, 730 ("TCP stack %s requested tcp_drop(%p)", 731 tp->t_fb->tfb_tcp_block_name, tp)); 732 return (rv); 733 } 734 735 /* 736 * tcp_unlock_or_drop() 737 * Handle return code from tfb_tcp_output() after we have logged/traced, 738 * to be used with tcp_output_nodrop(). 739 */ 740 static inline int 741 tcp_unlock_or_drop(struct tcpcb *tp, int tcp_output_retval) 742 { 743 struct inpcb *inp = tptoinpcb(tp); 744 745 INP_WLOCK_ASSERT(inp); 746 747 if (tcp_output_retval < 0) { 748 tcp_output_retval = -tcp_output_retval; 749 if (tcp_drop(tp, tcp_output_retval) != NULL) 750 INP_WUNLOCK(inp); 751 } else 752 INP_WUNLOCK(inp); 753 754 return (tcp_output_retval); 755 } 756 #endif /* _NETINET_IN_PCB_H_ */ 757 758 static int inline 759 tcp_packets_this_ack(struct tcpcb *tp, tcp_seq ack) 760 { 761 return ((ack - tp->snd_una) / tp->t_maxseg + 762 ((((ack - tp->snd_una) % tp->t_maxseg) != 0) ? 1 : 0)); 763 } 764 #endif /* _KERNEL */ 765 766 /* 767 * Flags and utility macros for the t_flags field. 768 */ 769 #define TF_ACKNOW 0x00000001 /* ack peer immediately */ 770 #define TF_DELACK 0x00000002 /* ack, but try to delay it */ 771 #define TF_NODELAY 0x00000004 /* don't delay packets to coalesce */ 772 #define TF_NOOPT 0x00000008 /* don't use tcp options */ 773 #define TF_SENTFIN 0x00000010 /* have sent FIN */ 774 #define TF_REQ_SCALE 0x00000020 /* have/will request window scaling */ 775 #define TF_RCVD_SCALE 0x00000040 /* other side has requested scaling */ 776 #define TF_REQ_TSTMP 0x00000080 /* have/will request timestamps */ 777 #define TF_RCVD_TSTMP 0x00000100 /* a timestamp was received in SYN */ 778 #define TF_SACK_PERMIT 0x00000200 /* other side said I could SACK */ 779 #define TF_NEEDSYN 0x00000400 /* send SYN (implicit state) */ 780 #define TF_NEEDFIN 0x00000800 /* send FIN (implicit state) */ 781 #define TF_NOPUSH 0x00001000 /* don't push */ 782 #define TF_PREVVALID 0x00002000 /* saved values for bad rxmit valid 783 * Note: accessing and restoring from 784 * these may only be done in the 1st 785 * RTO recovery round (t_rxtshift == 1) 786 */ 787 #define TF_WAKESOR 0x00004000 /* wake up receive socket */ 788 #define TF_GPUTINPROG 0x00008000 /* Goodput measurement in progress */ 789 #define TF_MORETOCOME 0x00010000 /* More data to be appended to sock */ 790 #define TF_SONOTCONN 0x00020000 /* needs soisconnected() on ESTAB */ 791 #define TF_LASTIDLE 0x00040000 /* connection was previously idle */ 792 #define TF_RXWIN0SENT 0x00080000 /* sent a receiver win 0 in response */ 793 #define TF_FASTRECOVERY 0x00100000 /* in NewReno Fast Recovery */ 794 #define TF_WASFRECOVERY 0x00200000 /* was in NewReno Fast Recovery */ 795 #define TF_SIGNATURE 0x00400000 /* require MD5 digests (RFC2385) */ 796 #define TF_FORCEDATA 0x00800000 /* force out a byte */ 797 #define TF_TSO 0x01000000 /* TSO enabled on this connection */ 798 #define TF_TOE 0x02000000 /* this connection is offloaded */ 799 #define TF_CLOSED 0x04000000 /* close(2) called on socket */ 800 #define TF_UNUSED1 0x08000000 /* unused */ 801 #define TF_LRD 0x10000000 /* Lost Retransmission Detection */ 802 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */ 803 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */ 804 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */ 805 806 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY) 807 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY 808 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY 809 810 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY) 811 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY 812 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY 813 814 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY)) 815 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY) 816 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY) 817 818 #if defined(_KERNEL) && !defined(TCP_RFC7413) 819 #define IS_FASTOPEN(t_flags) (false) 820 #else 821 #define IS_FASTOPEN(t_flags) (t_flags & TF_FASTOPEN) 822 #endif 823 824 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una) 825 826 /* 827 * Flags for the t_oobflags field. 828 */ 829 #define TCPOOB_HAVEDATA 0x01 830 #define TCPOOB_HADDATA 0x02 831 832 /* 833 * Flags for the extended TCP flags field, t_flags2 834 */ 835 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */ 836 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */ 837 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */ 838 #define TF2_LOG_AUTO 0x00000008 /* Session is auto-logging. */ 839 #define TF2_DROP_AF_DATA 0x00000010 /* Drop after all data ack'd */ 840 #define TF2_ECN_PERMIT 0x00000020 /* connection ECN-ready */ 841 #define TF2_ECN_SND_CWR 0x00000040 /* ECN CWR in queue */ 842 #define TF2_ECN_SND_ECE 0x00000080 /* ECN ECE in queue */ 843 #define TF2_ACE_PERMIT 0x00000100 /* Accurate ECN mode */ 844 #define TF2_FBYTES_COMPLETE 0x00000400 /* We have first bytes in and out */ 845 #define TF2_ECN_USE_ECT1 0x00000800 /* Use ECT(1) marking on session */ 846 #define TF2_TCP_ACCOUNTING 0x00010000 /* Do TCP accounting */ 847 848 /* 849 * Structure to hold TCP options that are only used during segment 850 * processing (in tcp_input), but not held in the tcpcb. 851 * It's basically used to reduce the number of parameters 852 * to tcp_dooptions and tcp_addoptions. 853 * The binary order of the to_flags is relevant for packing of the 854 * options in tcp_addoptions. 855 */ 856 struct tcpopt { 857 u_int32_t to_flags; /* which options are present */ 858 #define TOF_MSS 0x0001 /* maximum segment size */ 859 #define TOF_SCALE 0x0002 /* window scaling */ 860 #define TOF_SACKPERM 0x0004 /* SACK permitted */ 861 #define TOF_TS 0x0010 /* timestamp */ 862 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */ 863 #define TOF_SACK 0x0080 /* Peer sent SACK option */ 864 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */ 865 #define TOF_MAXOPT 0x0200 866 u_int32_t to_tsval; /* new timestamp */ 867 u_int32_t to_tsecr; /* reflected timestamp */ 868 u_char *to_sacks; /* pointer to the first SACK blocks */ 869 u_char *to_signature; /* pointer to the TCP-MD5 signature */ 870 u_int8_t *to_tfo_cookie; /* pointer to the TFO cookie */ 871 u_int16_t to_mss; /* maximum segment size */ 872 u_int8_t to_wscale; /* window scaling */ 873 u_int8_t to_nsacks; /* number of SACK blocks */ 874 u_int8_t to_tfo_len; /* TFO cookie length */ 875 u_int32_t to_spare; /* UTO */ 876 }; 877 878 /* 879 * Flags for tcp_dooptions. 880 */ 881 #define TO_SYN 0x01 /* parse SYN-only options */ 882 883 struct hc_metrics_lite { /* must stay in sync with hc_metrics */ 884 uint32_t rmx_mtu; /* MTU for this path */ 885 uint32_t rmx_ssthresh; /* outbound gateway buffer limit */ 886 uint32_t rmx_rtt; /* estimated round trip time */ 887 uint32_t rmx_rttvar; /* estimated rtt variance */ 888 uint32_t rmx_cwnd; /* congestion window */ 889 uint32_t rmx_sendpipe; /* outbound delay-bandwidth product */ 890 uint32_t rmx_recvpipe; /* inbound delay-bandwidth product */ 891 }; 892 893 /* 894 * Used by tcp_maxmtu() to communicate interface specific features 895 * and limits at the time of connection setup. 896 */ 897 struct tcp_ifcap { 898 int ifcap; 899 u_int tsomax; 900 u_int tsomaxsegcount; 901 u_int tsomaxsegsize; 902 }; 903 904 #ifndef _NETINET_IN_PCB_H_ 905 struct in_conninfo; 906 #endif /* _NETINET_IN_PCB_H_ */ 907 908 /* 909 * The smoothed round-trip time and estimated variance 910 * are stored as fixed point numbers scaled by the values below. 911 * For convenience, these scales are also used in smoothing the average 912 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). 913 * With these scales, srtt has 3 bits to the right of the binary point, 914 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the 915 * binary point, and is smoothed with an ALPHA of 0.75. 916 */ 917 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ 918 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ 919 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ 920 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ 921 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ 922 923 /* 924 * The initial retransmission should happen at rtt + 4 * rttvar. 925 * Because of the way we do the smoothing, srtt and rttvar 926 * will each average +1/2 tick of bias. When we compute 927 * the retransmit timer, we want 1/2 tick of rounding and 928 * 1 extra tick because of +-1/2 tick uncertainty in the 929 * firing of the timer. The bias will give us exactly the 930 * 1.5 tick we need. But, because the bias is 931 * statistical, we have to test that we don't drop below 932 * the minimum feasible timer (which is 2 ticks). 933 * This version of the macro adapted from a paper by Lawrence 934 * Brakmo and Larry Peterson which outlines a problem caused 935 * by insufficient precision in the original implementation, 936 * which results in inappropriately large RTO values for very 937 * fast networks. 938 */ 939 #define TCP_REXMTVAL(tp) \ 940 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ 941 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) 942 943 /* 944 * TCP statistics. 945 * Many of these should be kept per connection, 946 * but that's inconvenient at the moment. 947 */ 948 struct tcpstat { 949 uint64_t tcps_connattempt; /* connections initiated */ 950 uint64_t tcps_accepts; /* connections accepted */ 951 uint64_t tcps_connects; /* connections established */ 952 uint64_t tcps_drops; /* connections dropped */ 953 uint64_t tcps_conndrops; /* embryonic connections dropped */ 954 uint64_t tcps_minmssdrops; /* average minmss too low drops */ 955 uint64_t tcps_closed; /* conn. closed (includes drops) */ 956 uint64_t tcps_segstimed; /* segs where we tried to get rtt */ 957 uint64_t tcps_rttupdated; /* times we succeeded */ 958 uint64_t tcps_delack; /* delayed acks sent */ 959 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ 960 uint64_t tcps_rexmttimeo; /* retransmit timeouts */ 961 uint64_t tcps_persisttimeo; /* persist timeouts */ 962 uint64_t tcps_keeptimeo; /* keepalive timeouts */ 963 uint64_t tcps_keepprobe; /* keepalive probes sent */ 964 uint64_t tcps_keepdrops; /* connections dropped in keepalive */ 965 uint64_t tcps_progdrops; /* drops due to no progress */ 966 967 uint64_t tcps_sndtotal; /* total packets sent */ 968 uint64_t tcps_sndpack; /* data packets sent */ 969 uint64_t tcps_sndbyte; /* data bytes sent */ 970 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */ 971 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */ 972 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */ 973 uint64_t tcps_sndacks; /* ack-only packets sent */ 974 uint64_t tcps_sndprobe; /* window probes sent */ 975 uint64_t tcps_sndurg; /* packets sent with URG only */ 976 uint64_t tcps_sndwinup; /* window update-only packets sent */ 977 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ 978 979 uint64_t tcps_rcvtotal; /* total packets received */ 980 uint64_t tcps_rcvpack; /* packets received in sequence */ 981 uint64_t tcps_rcvbyte; /* bytes received in sequence */ 982 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */ 983 uint64_t tcps_rcvbadoff; /* packets received with bad offset */ 984 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */ 985 uint64_t tcps_rcvshort; /* packets received too short */ 986 uint64_t tcps_rcvduppack; /* duplicate-only packets received */ 987 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */ 988 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */ 989 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ 990 uint64_t tcps_rcvoopack; /* out-of-order packets received */ 991 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */ 992 uint64_t tcps_rcvpackafterwin; /* packets with data after window */ 993 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */ 994 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */ 995 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */ 996 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */ 997 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */ 998 uint64_t tcps_rcvackpack; /* rcvd ack packets */ 999 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */ 1000 uint64_t tcps_rcvwinupd; /* rcvd window update packets */ 1001 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */ 1002 uint64_t tcps_predack; /* times hdr predict ok for acks */ 1003 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */ 1004 uint64_t tcps_pcbcachemiss; 1005 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */ 1006 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */ 1007 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */ 1008 uint64_t tcps_usedrtt; /* times RTT initialized from route */ 1009 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */ 1010 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/ 1011 uint64_t tcps_persistdrop; /* timeout in persist state */ 1012 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */ 1013 uint64_t tcps_mturesent; /* resends due to MTU discovery */ 1014 uint64_t tcps_listendrop; /* listen queue overflows */ 1015 uint64_t tcps_badrst; /* ignored RSTs in the window */ 1016 1017 uint64_t tcps_sc_added; /* entry added to syncache */ 1018 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */ 1019 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */ 1020 uint64_t tcps_sc_dropped; /* could not reply to packet */ 1021 uint64_t tcps_sc_completed; /* successful extraction of entry */ 1022 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */ 1023 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */ 1024 uint64_t tcps_sc_reset; /* RST removed entry from syncache */ 1025 uint64_t tcps_sc_stale; /* timed out or listen socket gone */ 1026 uint64_t tcps_sc_aborted; /* syncache entry aborted */ 1027 uint64_t tcps_sc_badack; /* removed due to bad ACK */ 1028 uint64_t tcps_sc_unreach; /* ICMP unreachable received */ 1029 uint64_t tcps_sc_zonefail; /* zalloc() failed */ 1030 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */ 1031 uint64_t tcps_sc_recvcookie; /* SYN cookie received */ 1032 1033 uint64_t tcps_hc_added; /* entry added to hostcache */ 1034 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */ 1035 1036 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ 1037 1038 /* SACK related stats */ 1039 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */ 1040 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */ 1041 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ 1042 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ 1043 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ 1044 uint64_t tcps_sack_lostrexmt; /* SACK lost retransmission recovered */ 1045 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */ 1046 1047 /* ECN related stats */ 1048 uint64_t tcps_ecn_rcvce; /* ECN Congestion Experienced */ 1049 uint64_t tcps_ecn_rcvect0; /* ECN Capable Transport */ 1050 uint64_t tcps_ecn_rcvect1; /* ECN Capable Transport */ 1051 uint64_t tcps_ecn_shs; /* ECN successful handshakes */ 1052 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */ 1053 1054 /* TCP_SIGNATURE related stats */ 1055 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */ 1056 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */ 1057 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */ 1058 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */ 1059 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */ 1060 1061 /* Path MTU Discovery Black Hole Detection related stats */ 1062 uint64_t tcps_pmtud_blackhole_activated; /* Black Hole Count */ 1063 uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */ 1064 uint64_t tcps_pmtud_blackhole_failed; /* Black Hole Failure Count */ 1065 1066 uint64_t tcps_tunneled_pkts; /* Packets encap's in UDP received */ 1067 uint64_t tcps_tunneled_errs; /* Packets that had errors that were UDP encaped */ 1068 1069 /* Dsack related stats */ 1070 uint64_t tcps_dsack_count; /* Number of ACKs arriving with DSACKs */ 1071 uint64_t tcps_dsack_bytes; /* Number of bytes DSACK'ed no TLP */ 1072 uint64_t tcps_dsack_tlp_bytes; /* Number of bytes DSACK'ed due to TLPs */ 1073 1074 /* TCPS_TIME_WAIT usage stats */ 1075 uint64_t tcps_tw_recycles; /* Times time-wait was recycled. */ 1076 uint64_t tcps_tw_resets; /* Times time-wait sent a reset. */ 1077 uint64_t tcps_tw_responds; /* Times time-wait sent a valid ack. */ 1078 1079 /* Accurate ECN Handshake stats */ 1080 uint64_t tcps_ace_nect; /* ACE SYN packet with Non-ECT */ 1081 uint64_t tcps_ace_ect1; /* ACE SYN packet with ECT1 */ 1082 uint64_t tcps_ace_ect0; /* ACE SYN packet with ECT0 */ 1083 uint64_t tcps_ace_ce; /* ACE SYN packet with CE */ 1084 1085 /* ECN related stats */ 1086 uint64_t tcps_ecn_sndect0; /* ECN Capable Transport */ 1087 uint64_t tcps_ecn_sndect1; /* ECN Capable Transport */ 1088 1089 uint64_t _pad[4]; /* 4 TBD placeholder for STABLE */ 1090 }; 1091 1092 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */ 1093 1094 #ifdef _KERNEL 1095 #define TI_UNLOCKED 1 1096 #define TI_RLOCKED 2 1097 #include <sys/counter.h> 1098 1099 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */ 1100 /* 1101 * In-kernel consumers can use these accessor macros directly to update 1102 * stats. 1103 */ 1104 #define TCPSTAT_ADD(name, val) \ 1105 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)) 1106 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1) 1107 1108 /* 1109 * Kernel module consumers must use this accessor macro. 1110 */ 1111 void kmod_tcpstat_add(int statnum, int val); 1112 #define KMOD_TCPSTAT_ADD(name, val) \ 1113 kmod_tcpstat_add(offsetof(struct tcpstat, name) / sizeof(uint64_t), val) 1114 #define KMOD_TCPSTAT_INC(name) KMOD_TCPSTAT_ADD(name, 1) 1115 1116 /* 1117 * Running TCP connection count by state. 1118 */ 1119 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]); 1120 #define V_tcps_states VNET(tcps_states) 1121 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1) 1122 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1) 1123 1124 /* 1125 * TCP specific helper hook point identifiers. 1126 */ 1127 #define HHOOK_TCP_EST_IN 0 1128 #define HHOOK_TCP_EST_OUT 1 1129 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT 1130 1131 struct tcp_hhook_data { 1132 struct tcpcb *tp; 1133 struct tcphdr *th; 1134 struct tcpopt *to; 1135 uint32_t len; 1136 int tso; 1137 tcp_seq curack; 1138 }; 1139 #ifdef TCP_HHOOK 1140 void hhook_run_tcp_est_out(struct tcpcb *tp, 1141 struct tcphdr *th, struct tcpopt *to, 1142 uint32_t len, int tso); 1143 #endif 1144 #endif 1145 1146 /* 1147 * TCB structure exported to user-land via sysctl(3). 1148 * 1149 * Fields prefixed with "xt_" are unique to the export structure, and fields 1150 * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'. 1151 * 1152 * Legend: 1153 * (s) - used by userland utilities in src 1154 * (p) - used by utilities in ports 1155 * (3) - is known to be used by third party software not in ports 1156 * (n) - no known usage 1157 * 1158 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been 1159 * included. Not all of our clients do. 1160 */ 1161 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) 1162 struct xtcpcb { 1163 ksize_t xt_len; /* length of this structure */ 1164 struct xinpcb xt_inp; 1165 char xt_stack[TCP_FUNCTION_NAME_LEN_MAX]; /* (s) */ 1166 char xt_logid[TCP_LOG_ID_LEN]; /* (s) */ 1167 char xt_cc[TCP_CA_NAME_MAX]; /* (s) */ 1168 int64_t spare64[6]; 1169 int32_t t_state; /* (s,p) */ 1170 uint32_t t_flags; /* (s,p) */ 1171 int32_t t_sndzerowin; /* (s) */ 1172 int32_t t_sndrexmitpack; /* (s) */ 1173 int32_t t_rcvoopack; /* (s) */ 1174 int32_t t_rcvtime; /* (s) */ 1175 int32_t tt_rexmt; /* (s) */ 1176 int32_t tt_persist; /* (s) */ 1177 int32_t tt_keep; /* (s) */ 1178 int32_t tt_2msl; /* (s) */ 1179 int32_t tt_delack; /* (s) */ 1180 int32_t t_logstate; /* (3) */ 1181 uint32_t t_snd_cwnd; /* (s) */ 1182 uint32_t t_snd_ssthresh; /* (s) */ 1183 uint32_t t_maxseg; /* (s) */ 1184 uint32_t t_rcv_wnd; /* (s) */ 1185 uint32_t t_snd_wnd; /* (s) */ 1186 uint32_t xt_ecn; /* (s) */ 1187 uint32_t t_dsack_bytes; /* (n) */ 1188 uint32_t t_dsack_tlp_bytes; /* (n) */ 1189 uint32_t t_dsack_pack; /* (n) */ 1190 uint16_t xt_encaps_port; /* (s) */ 1191 int16_t spare16; 1192 int32_t spare32[22]; 1193 } __aligned(8); 1194 1195 #ifdef _KERNEL 1196 void tcp_inptoxtp(const struct inpcb *, struct xtcpcb *); 1197 #endif 1198 #endif 1199 1200 /* 1201 * TCP function information (name-to-id mapping, aliases, and refcnt) 1202 * exported to user-land via sysctl(3). 1203 */ 1204 struct tcp_function_info { 1205 uint32_t tfi_refcnt; 1206 uint8_t tfi_id; 1207 char tfi_name[TCP_FUNCTION_NAME_LEN_MAX]; 1208 char tfi_alias[TCP_FUNCTION_NAME_LEN_MAX]; 1209 }; 1210 1211 /* 1212 * Identifiers for TCP sysctl nodes 1213 */ 1214 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ 1215 #define TCPCTL_MSSDFLT 3 /* MSS default */ 1216 #define TCPCTL_STATS 4 /* statistics */ 1217 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */ 1218 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ 1219 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ 1220 #define TCPCTL_SENDSPACE 8 /* send buffer space */ 1221 #define TCPCTL_RECVSPACE 9 /* receive buffer space */ 1222 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ 1223 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ 1224 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ 1225 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ 1226 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ 1227 #define TCPCTL_DROP 15 /* drop tcp connection */ 1228 #define TCPCTL_STATES 16 /* connection counts by TCP state */ 1229 1230 #ifdef _KERNEL 1231 #ifdef SYSCTL_DECL 1232 SYSCTL_DECL(_net_inet_tcp); 1233 SYSCTL_DECL(_net_inet_tcp_sack); 1234 MALLOC_DECLARE(M_TCPLOG); 1235 #endif 1236 1237 VNET_DECLARE(int, tcp_log_in_vain); 1238 #define V_tcp_log_in_vain VNET(tcp_log_in_vain) 1239 1240 /* 1241 * Global TCP tunables shared between different stacks. 1242 * Please keep the list sorted. 1243 */ 1244 VNET_DECLARE(int, drop_synfin); 1245 VNET_DECLARE(int, path_mtu_discovery); 1246 VNET_DECLARE(int, tcp_abc_l_var); 1247 VNET_DECLARE(int, tcp_autorcvbuf_max); 1248 VNET_DECLARE(int, tcp_autosndbuf_inc); 1249 VNET_DECLARE(int, tcp_autosndbuf_max); 1250 VNET_DECLARE(int, tcp_delack_enabled); 1251 VNET_DECLARE(int, tcp_do_autorcvbuf); 1252 VNET_DECLARE(int, tcp_do_autosndbuf); 1253 VNET_DECLARE(int, tcp_do_ecn); 1254 VNET_DECLARE(int, tcp_do_lrd); 1255 VNET_DECLARE(int, tcp_do_prr); 1256 VNET_DECLARE(int, tcp_do_prr_conservative); 1257 VNET_DECLARE(int, tcp_do_newcwv); 1258 VNET_DECLARE(int, tcp_do_rfc1323); 1259 VNET_DECLARE(int, tcp_tolerate_missing_ts); 1260 VNET_DECLARE(int, tcp_do_rfc3042); 1261 VNET_DECLARE(int, tcp_do_rfc3390); 1262 VNET_DECLARE(int, tcp_do_rfc3465); 1263 VNET_DECLARE(int, tcp_do_newsack); 1264 VNET_DECLARE(int, tcp_do_sack); 1265 VNET_DECLARE(int, tcp_do_tso); 1266 VNET_DECLARE(int, tcp_ecn_maxretries); 1267 VNET_DECLARE(int, tcp_initcwnd_segments); 1268 VNET_DECLARE(int, tcp_insecure_rst); 1269 VNET_DECLARE(int, tcp_insecure_syn); 1270 VNET_DECLARE(uint32_t, tcp_map_entries_limit); 1271 VNET_DECLARE(uint32_t, tcp_map_split_limit); 1272 VNET_DECLARE(int, tcp_minmss); 1273 VNET_DECLARE(int, tcp_mssdflt); 1274 #ifdef STATS 1275 VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl); 1276 VNET_DECLARE(int, tcp_perconn_stats_enable); 1277 #endif /* STATS */ 1278 VNET_DECLARE(int, tcp_recvspace); 1279 VNET_DECLARE(int, tcp_sack_globalholes); 1280 VNET_DECLARE(int, tcp_sack_globalmaxholes); 1281 VNET_DECLARE(int, tcp_sack_maxholes); 1282 VNET_DECLARE(int, tcp_sc_rst_sock_fail); 1283 VNET_DECLARE(int, tcp_sendspace); 1284 VNET_DECLARE(int, tcp_udp_tunneling_overhead); 1285 VNET_DECLARE(int, tcp_udp_tunneling_port); 1286 VNET_DECLARE(struct inpcbinfo, tcbinfo); 1287 1288 #define V_tcp_do_lrd VNET(tcp_do_lrd) 1289 #define V_tcp_do_prr VNET(tcp_do_prr) 1290 #define V_tcp_do_prr_conservative VNET(tcp_do_prr_conservative) 1291 #define V_tcp_do_newcwv VNET(tcp_do_newcwv) 1292 #define V_drop_synfin VNET(drop_synfin) 1293 #define V_path_mtu_discovery VNET(path_mtu_discovery) 1294 #define V_tcbinfo VNET(tcbinfo) 1295 #define V_tcp_abc_l_var VNET(tcp_abc_l_var) 1296 #define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max) 1297 #define V_tcp_autosndbuf_inc VNET(tcp_autosndbuf_inc) 1298 #define V_tcp_autosndbuf_max VNET(tcp_autosndbuf_max) 1299 #define V_tcp_delack_enabled VNET(tcp_delack_enabled) 1300 #define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf) 1301 #define V_tcp_do_autosndbuf VNET(tcp_do_autosndbuf) 1302 #define V_tcp_do_ecn VNET(tcp_do_ecn) 1303 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323) 1304 #define V_tcp_tolerate_missing_ts VNET(tcp_tolerate_missing_ts) 1305 #define V_tcp_ts_offset_per_conn VNET(tcp_ts_offset_per_conn) 1306 #define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042) 1307 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390) 1308 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465) 1309 #define V_tcp_do_newsack VNET(tcp_do_newsack) 1310 #define V_tcp_do_sack VNET(tcp_do_sack) 1311 #define V_tcp_do_tso VNET(tcp_do_tso) 1312 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries) 1313 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments) 1314 #define V_tcp_insecure_rst VNET(tcp_insecure_rst) 1315 #define V_tcp_insecure_syn VNET(tcp_insecure_syn) 1316 #define V_tcp_map_entries_limit VNET(tcp_map_entries_limit) 1317 #define V_tcp_map_split_limit VNET(tcp_map_split_limit) 1318 #define V_tcp_minmss VNET(tcp_minmss) 1319 #define V_tcp_mssdflt VNET(tcp_mssdflt) 1320 #ifdef STATS 1321 #define V_tcp_perconn_stats_dflt_tpl VNET(tcp_perconn_stats_dflt_tpl) 1322 #define V_tcp_perconn_stats_enable VNET(tcp_perconn_stats_enable) 1323 #endif /* STATS */ 1324 #define V_tcp_recvspace VNET(tcp_recvspace) 1325 #define V_tcp_sack_globalholes VNET(tcp_sack_globalholes) 1326 #define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes) 1327 #define V_tcp_sack_maxholes VNET(tcp_sack_maxholes) 1328 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail) 1329 #define V_tcp_sendspace VNET(tcp_sendspace) 1330 #define V_tcp_udp_tunneling_overhead VNET(tcp_udp_tunneling_overhead) 1331 #define V_tcp_udp_tunneling_port VNET(tcp_udp_tunneling_port) 1332 1333 #ifdef TCP_HHOOK 1334 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]); 1335 #define V_tcp_hhh VNET(tcp_hhh) 1336 #endif 1337 1338 int tcp_addoptions(struct tcpopt *, u_char *); 1339 struct tcpcb * 1340 tcp_close(struct tcpcb *); 1341 void tcp_discardcb(struct tcpcb *); 1342 void tcp_twstart(struct tcpcb *); 1343 int tcp_ctloutput(struct socket *, struct sockopt *); 1344 void tcp_fini(void *); 1345 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, const void *, 1346 const void *); 1347 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, const void *, 1348 const void *); 1349 int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *, 1350 struct mbuf *); 1351 void tcp_reass_global_init(void); 1352 void tcp_reass_flush(struct tcpcb *); 1353 void tcp_dooptions(struct tcpopt *, u_char *, int, int); 1354 void tcp_dropwithreset(struct mbuf *, struct tcphdr *, 1355 struct tcpcb *, int, int); 1356 void tcp_pulloutofband(struct socket *, 1357 struct tcphdr *, struct mbuf *, int); 1358 void tcp_xmit_timer(struct tcpcb *, int); 1359 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); 1360 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th, 1361 uint16_t nsegs, uint16_t type); 1362 void cc_conn_init(struct tcpcb *tp); 1363 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); 1364 void cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos); 1365 void cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos); 1366 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type); 1367 #ifdef TCP_HHOOK 1368 void hhook_run_tcp_est_in(struct tcpcb *tp, 1369 struct tcphdr *th, struct tcpopt *to); 1370 #endif 1371 1372 int tcp_input(struct mbuf **, int *, int); 1373 int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *, 1374 struct tcpcb *, int); 1375 int tcp_input_with_port(struct mbuf **, int *, int, uint16_t); 1376 void tcp_do_segment(struct mbuf *, struct tcphdr *, 1377 struct socket *, struct tcpcb *, int, int, uint8_t); 1378 1379 int register_tcp_functions(struct tcp_function_block *blk, int wait); 1380 int register_tcp_functions_as_names(struct tcp_function_block *blk, 1381 int wait, const char *names[], int *num_names); 1382 int register_tcp_functions_as_name(struct tcp_function_block *blk, 1383 const char *name, int wait); 1384 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce, 1385 bool force); 1386 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs); 1387 int find_tcp_function_alias(struct tcp_function_block *blk, struct tcp_function_set *fs); 1388 void tcp_switch_back_to_default(struct tcpcb *tp); 1389 struct tcp_function_block * 1390 find_and_ref_tcp_fb(struct tcp_function_block *fs); 1391 int tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt); 1392 int tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt); 1393 1394 extern counter_u64_t tcp_inp_lro_direct_queue; 1395 extern counter_u64_t tcp_inp_lro_wokeup_queue; 1396 extern counter_u64_t tcp_inp_lro_compressed; 1397 extern counter_u64_t tcp_inp_lro_locks_taken; 1398 extern counter_u64_t tcp_extra_mbuf; 1399 extern counter_u64_t tcp_would_have_but; 1400 extern counter_u64_t tcp_comp_total; 1401 extern counter_u64_t tcp_uncomp_total; 1402 extern counter_u64_t tcp_bad_csums; 1403 1404 #ifdef NETFLIX_EXP_DETECTION 1405 /* Various SACK attack thresholds */ 1406 extern int32_t tcp_force_detection; 1407 extern int32_t tcp_sad_limit; 1408 extern int32_t tcp_sack_to_ack_thresh; 1409 extern int32_t tcp_sack_to_move_thresh; 1410 extern int32_t tcp_restoral_thresh; 1411 extern int32_t tcp_sad_decay_val; 1412 extern int32_t tcp_sad_pacing_interval; 1413 extern int32_t tcp_sad_low_pps; 1414 extern int32_t tcp_map_minimum; 1415 extern int32_t tcp_attack_on_turns_on_logging; 1416 #endif 1417 extern uint32_t tcp_ack_war_time_window; 1418 extern uint32_t tcp_ack_war_cnt; 1419 1420 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *); 1421 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *); 1422 void tcp6_use_min_mtu(struct tcpcb *); 1423 u_int tcp_maxseg(const struct tcpcb *); 1424 u_int tcp_fixed_maxseg(const struct tcpcb *); 1425 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *, 1426 struct tcp_ifcap *); 1427 void tcp_mss(struct tcpcb *, int); 1428 int tcp_mssopt(struct in_conninfo *); 1429 struct tcpcb * 1430 tcp_newtcpcb(struct inpcb *); 1431 int tcp_default_output(struct tcpcb *); 1432 void tcp_state_change(struct tcpcb *, int); 1433 void tcp_respond(struct tcpcb *, void *, 1434 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, uint16_t); 1435 bool tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *, 1436 struct mbuf *, int); 1437 void tcp_setpersist(struct tcpcb *); 1438 void tcp_record_dsack(struct tcpcb *tp, tcp_seq start, tcp_seq end, int tlp); 1439 struct tcptemp * 1440 tcpip_maketemplate(struct inpcb *); 1441 void tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *); 1442 void tcp_timer_activate(struct tcpcb *, tt_which, u_int); 1443 bool tcp_timer_active(struct tcpcb *, tt_which); 1444 void tcp_timer_stop(struct tcpcb *); 1445 int inp_to_cpuid(struct inpcb *inp); 1446 /* 1447 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) 1448 */ 1449 void tcp_hc_init(void); 1450 #ifdef VIMAGE 1451 void tcp_hc_destroy(void); 1452 #endif 1453 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); 1454 uint32_t tcp_hc_getmtu(struct in_conninfo *); 1455 void tcp_hc_updatemtu(struct in_conninfo *, uint32_t); 1456 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); 1457 void cc_after_idle(struct tcpcb *tp); 1458 1459 extern struct protosw tcp_protosw; /* shared for TOE */ 1460 extern struct protosw tcp6_protosw; /* shared for TOE */ 1461 1462 uint32_t tcp_new_ts_offset(struct in_conninfo *); 1463 tcp_seq tcp_new_isn(struct in_conninfo *); 1464 1465 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); 1466 int tcp_dsack_block_exists(struct tcpcb *); 1467 void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq); 1468 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); 1469 void tcp_clean_dsack_blocks(struct tcpcb *tp); 1470 void tcp_clean_sackreport(struct tcpcb *tp); 1471 void tcp_sack_adjust(struct tcpcb *tp); 1472 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); 1473 void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *); 1474 void tcp_lost_retransmission(struct tcpcb *, struct tcphdr *); 1475 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); 1476 void tcp_free_sackholes(struct tcpcb *tp); 1477 void tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *); 1478 int tcp_newreno(struct tcpcb *, struct tcphdr *); 1479 int tcp_compute_pipe(struct tcpcb *); 1480 uint32_t tcp_compute_initwnd(uint32_t); 1481 void tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t); 1482 int tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes, 1483 size_t seed_len); 1484 int tcp_can_enable_pacing(void); 1485 void tcp_decrement_paced_conn(void); 1486 void tcp_change_time_units(struct tcpcb *, int); 1487 void tcp_handle_orphaned_packets(struct tcpcb *); 1488 1489 struct mbuf * 1490 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen, 1491 int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls); 1492 1493 int tcp_stats_init(void); 1494 void tcp_log_end_status(struct tcpcb *tp, uint8_t status); 1495 #ifdef TCP_REQUEST_TRK 1496 void tcp_http_free_a_slot(struct tcpcb *tp, struct http_sendfile_track *ent); 1497 struct http_sendfile_track * 1498 tcp_http_find_a_req_that_is_completed_by(struct tcpcb *tp, tcp_seq th_ack, int *ip); 1499 int tcp_http_check_for_comp(struct tcpcb *tp, tcp_seq ack_point); 1500 int 1501 tcp_http_is_entry_comp(struct tcpcb *tp, struct http_sendfile_track *ent, tcp_seq ack_point); 1502 struct http_sendfile_track * 1503 tcp_http_find_req_for_seq(struct tcpcb *tp, tcp_seq seq); 1504 void 1505 tcp_http_log_req_info(struct tcpcb *tp, 1506 struct http_sendfile_track *http, uint16_t slot, 1507 uint8_t val, uint64_t offset, uint64_t nbytes); 1508 1509 uint32_t 1510 tcp_estimate_tls_overhead(struct socket *so, uint64_t tls_usr_bytes); 1511 void 1512 tcp_http_alloc_req(struct tcpcb *tp, union tcp_log_userdata *user, 1513 uint64_t ts); 1514 1515 struct http_sendfile_track * 1516 tcp_http_alloc_req_full(struct tcpcb *tp, struct http_req *req, uint64_t ts, int rec_dups); 1517 1518 1519 #endif 1520 #ifdef TCP_ACCOUNTING 1521 int tcp_do_ack_accounting(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to, uint32_t tiwin, int mss); 1522 #endif 1523 1524 1525 static inline void 1526 tcp_fields_to_host(struct tcphdr *th) 1527 { 1528 1529 th->th_seq = ntohl(th->th_seq); 1530 th->th_ack = ntohl(th->th_ack); 1531 th->th_win = ntohs(th->th_win); 1532 th->th_urp = ntohs(th->th_urp); 1533 } 1534 1535 static inline void 1536 tcp_fields_to_net(struct tcphdr *th) 1537 { 1538 1539 th->th_seq = htonl(th->th_seq); 1540 th->th_ack = htonl(th->th_ack); 1541 th->th_win = htons(th->th_win); 1542 th->th_urp = htons(th->th_urp); 1543 } 1544 1545 static inline uint16_t 1546 tcp_get_flags(const struct tcphdr *th) 1547 { 1548 return (((uint16_t)th->th_x2 << 8) | th->th_flags); 1549 } 1550 1551 static inline void 1552 tcp_set_flags(struct tcphdr *th, uint16_t flags) 1553 { 1554 th->th_x2 = (flags >> 8) & 0x0f; 1555 th->th_flags = flags & 0xff; 1556 } 1557 1558 static inline void 1559 tcp_account_for_send(struct tcpcb *tp, uint32_t len, uint8_t is_rxt, 1560 uint8_t is_tlp, bool hw_tls) 1561 { 1562 if (is_tlp) { 1563 tp->t_sndtlppack++; 1564 tp->t_sndtlpbyte += len; 1565 } 1566 /* To get total bytes sent you must add t_snd_rxt_bytes to t_sndbytes */ 1567 if (is_rxt) 1568 tp->t_snd_rxt_bytes += len; 1569 else 1570 tp->t_sndbytes += len; 1571 1572 #ifdef KERN_TLS 1573 if (hw_tls && is_rxt && len != 0) { 1574 uint64_t rexmit_percent = (1000ULL * tp->t_snd_rxt_bytes) / (10ULL * (tp->t_snd_rxt_bytes + tp->t_sndbytes)); 1575 if (rexmit_percent > ktls_ifnet_max_rexmit_pct) 1576 ktls_disable_ifnet(tp); 1577 } 1578 #endif 1579 1580 } 1581 #endif /* _KERNEL */ 1582 1583 #endif /* _NETINET_TCP_VAR_H_ */ 1584