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