2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1993, 1994, 1995
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95
35 #ifndef _NETINET_TCP_VAR_H_
36 #define _NETINET_TCP_VAR_H_
38 #include <netinet/tcp.h>
39 #include <netinet/tcp_fsm.h>
46 #if defined(_KERNEL) || defined(_WANT_TCPCB)
47 /* TCP segment queue entry */
49 TAILQ_ENTRY(tseg_qent) tqe_q;
50 struct mbuf *tqe_m; /* mbuf contains packet */
51 struct mbuf *tqe_last; /* last mbuf in chain */
52 tcp_seq tqe_start; /* TCP Sequence number start */
53 int tqe_len; /* TCP segment data length */
54 uint32_t tqe_flags; /* The flags from the th->th_flags */
55 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */
57 TAILQ_HEAD(tsegqe_head, tseg_qent);
60 tcp_seq start; /* start seq no. of sack block */
61 tcp_seq end; /* end seq no. */
65 tcp_seq start; /* start seq no. of hole */
66 tcp_seq end; /* end seq no. */
67 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
68 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
72 struct sackhole *nexthole;
73 int sack_bytes_rexmit;
74 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */
76 int ispare; /* explicit pad for 64bit alignment */
78 * Total sacked bytes reported by the
79 * receiver via sack option
81 uint32_t _pad1[1]; /* TBD */
82 uint64_t _pad[1]; /* TBD */
85 #define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
87 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
90 * Tcp control block, one per tcp; fields:
91 * Organized for 64 byte cacheline efficiency based
92 * on common tcp_input/tcp_output processing.
96 struct inpcb *t_inpcb; /* back pointer to internet pcb */
97 struct tcp_function_block *t_fb;/* TCP function call block */
98 void *t_fb_ptr; /* Pointer to t_fb specific data */
99 uint32_t t_maxseg:24, /* maximum segment size */
100 t_logstate:8; /* State of "black box" logging */
101 uint32_t t_port:16, /* Tunneling (over udp) port */
102 t_state:4, /* state of this connection */
104 t_delayed_ack: 7, /* Delayed ack variable */
107 tcp_seq snd_una; /* sent but unacknowledged */
108 tcp_seq snd_max; /* highest sequence number sent;
109 * used to recognize retransmits
111 tcp_seq snd_nxt; /* send next */
112 tcp_seq snd_up; /* send urgent pointer */
113 uint32_t snd_wnd; /* send window */
114 uint32_t snd_cwnd; /* congestion-controlled window */
115 uint32_t t_peakrate_thr; /* pre-calculated peak rate threshold */
117 u_int32_t ts_offset; /* our timestamp offset */
118 u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
119 int rcv_numsacks; /* # distinct sack blks present */
120 u_int t_tsomax; /* TSO total burst length limit in bytes */
121 u_int t_tsomaxsegcount; /* TSO maximum segment count */
122 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */
123 tcp_seq rcv_nxt; /* receive next */
124 tcp_seq rcv_adv; /* advertised window */
125 uint32_t rcv_wnd; /* receive window */
126 u_int t_flags2; /* More tcpcb flags storage */
127 int t_srtt; /* smoothed round-trip time */
128 int t_rttvar; /* variance in round-trip time */
129 u_int32_t ts_recent; /* timestamp echo data */
130 u_char snd_scale; /* window scaling for send window */
131 u_char rcv_scale; /* window scaling for recv window */
132 u_char snd_limited; /* segments limited transmitted */
133 u_char request_r_scale; /* pending window scaling */
134 tcp_seq last_ack_sent;
135 u_int t_rcvtime; /* inactivity time */
137 tcp_seq rcv_up; /* receive urgent pointer */
138 int t_segqlen; /* segment reassembly queue length */
139 uint32_t t_segqmbuflen; /* Count of bytes mbufs on all entries */
140 struct tsegqe_head t_segq; /* segment reassembly queue */
141 struct mbuf *t_in_pkt;
142 struct mbuf *t_tail_pkt;
143 struct tcp_timer *t_timers; /* All the TCP timers in one struct */
144 struct vnet *t_vnet; /* back pointer to parent vnet */
145 uint32_t snd_ssthresh; /* snd_cwnd size threshold for
146 * for slow start exponential to
149 tcp_seq snd_wl1; /* window update seg seq number */
151 tcp_seq snd_wl2; /* window update seg ack number */
153 tcp_seq irs; /* initial receive sequence number */
154 tcp_seq iss; /* initial send sequence number */
156 u_int ts_recent_age; /* when last updated */
157 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
158 uint16_t cl4_spare; /* Spare to adjust CL 4 */
159 char t_oobflags; /* have some */
160 char t_iobc; /* input character */
161 int t_rxtcur; /* current retransmit value (ticks) */
163 int t_rxtshift; /* log(2) of rexmt exp. backoff */
164 u_int t_rtttime; /* RTT measurement start time */
166 tcp_seq t_rtseq; /* sequence number being timed */
167 u_int t_starttime; /* time connection was established */
169 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */
170 u_int t_rttmin; /* minimum rtt allowed */
172 u_int t_rttbest; /* best rtt we've seen */
174 int t_softerror; /* possible error not yet reported */
175 uint32_t max_sndwnd; /* largest window peer has offered */
177 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */
178 uint32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
179 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
180 int t_sndzerowin; /* zero-window updates sent */
181 u_long t_rttupdated; /* number of times rtt sampled */
182 int snd_numholes; /* number of holes seen by sender */
183 u_int t_badrxtwin; /* window for retransmit recovery */
184 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
185 /* SACK scoreboard (sorted) */
186 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
187 tcp_seq sack_newdata; /* New data xmitted in this recovery
188 episode starts at this seq number */
189 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
190 struct sackhint sackhint; /* SACK scoreboard hint */
191 int t_rttlow; /* smallest observerved RTT */
192 int rfbuf_cnt; /* recv buffer autoscaling byte count */
193 struct toedev *tod; /* toedev handling this connection */
194 int t_sndrexmitpack; /* retransmit packets sent */
195 int t_rcvoopack; /* out-of-order packets received */
196 void *t_toe; /* TOE pcb pointer */
197 struct cc_algo *cc_algo; /* congestion control algorithm */
198 struct cc_var *ccv; /* congestion control specific vars */
199 struct osd *osd; /* storage for Khelp module data */
200 int t_bytes_acked; /* # bytes acked during current RTT */
201 u_int t_maxunacktime;
202 u_int t_keepinit; /* time to establish connection */
203 u_int t_keepidle; /* time before keepalive probes begin */
204 u_int t_keepintvl; /* interval between keepalives */
205 u_int t_keepcnt; /* number of keepalives before close */
206 int t_dupacks; /* consecutive dup acks recd */
207 int t_lognum; /* Number of log entries */
208 struct tcp_log_stailq t_logs; /* Log buffer */
209 struct tcp_log_id_node *t_lin;
210 struct tcp_log_id_bucket *t_lib;
211 const char *t_output_caller; /* Function that called tcp_output */
212 uint32_t t_logsn; /* Log "serial number" */
213 uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
214 unsigned int *t_tfo_pending; /* TCP Fast Open server pending counter */
216 uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
218 } t_tfo_cookie; /* TCP Fast Open cookie to send */
220 struct mbufq t_inpkts; /* List of saved input packets. */
221 struct mbufq t_outpkts; /* List of saved output packets. */
224 #endif /* _KERNEL || _WANT_TCPCB */
228 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
233 * TODO: We yet need to brave plowing in
234 * to tcp_input() and the pru_usrreq() block.
235 * Right now these go to the old standards which
236 * are somewhat ok, but in the long term may
237 * need to be changed. If we do tackle tcp_input()
238 * then we need to get rid of the tcp_do_segment()
241 /* Flags for tcp functions */
242 #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */
245 * If defining the optional tcp_timers, in the
246 * tfb_tcp_timer_stop call you must use the
247 * callout_async_drain() function with the
248 * tcp_timer_discard callback. You should check
249 * the return of callout_async_drain() and if 0
250 * increment tt_draincnt. Since the timer sub-system
251 * does not know your callbacks you must provide a
252 * stop_all function that loops through and calls
253 * tcp_timer_stop() with each of your defined timers.
254 * Adding a tfb_tcp_handoff_ok function allows the socket
255 * option to change stacks to query you even if the
256 * connection is in a later stage. You return 0 to
257 * say you can take over and run your stack, you return
258 * non-zero (an error number) to say no you can't.
259 * If the function is undefined you can only change
260 * in the early states (before connect or listen).
261 * tfb_tcp_fb_fini is changed to add a flag to tell
262 * the old stack if the tcb is being destroyed or
263 * not. A one in the flag means the TCB is being
264 * destroyed, a zero indicates its transitioning to
265 * another stack (via socket option).
267 struct tcp_function_block {
268 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
269 int (*tfb_tcp_output)(struct tcpcb *);
270 int (*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
271 void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
272 struct socket *, struct tcpcb *,
274 void (*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
275 struct socket *, struct tcpcb *,
277 int, struct timeval *);
278 int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
279 struct inpcb *inp, struct tcpcb *tp);
280 /* Optional memory allocation/free routine */
281 int (*tfb_tcp_fb_init)(struct tcpcb *);
282 void (*tfb_tcp_fb_fini)(struct tcpcb *, int);
283 /* Optional timers, must define all if you define one */
284 int (*tfb_tcp_timer_stop_all)(struct tcpcb *);
285 void (*tfb_tcp_timer_activate)(struct tcpcb *,
287 int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
288 void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
289 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *);
290 int (*tfb_tcp_handoff_ok)(struct tcpcb *);
291 void (*tfb_tcp_mtu_chg)(struct tcpcb *);
292 volatile uint32_t tfb_refcnt;
297 struct tcp_function {
298 TAILQ_ENTRY(tcp_function) tf_next;
299 char tf_name[TCP_FUNCTION_NAME_LEN_MAX];
300 struct tcp_function_block *tf_fb;
303 TAILQ_HEAD(tcp_funchead, tcp_function);
307 * Flags and utility macros for the t_flags field.
309 #define TF_ACKNOW 0x000001 /* ack peer immediately */
310 #define TF_DELACK 0x000002 /* ack, but try to delay it */
311 #define TF_NODELAY 0x000004 /* don't delay packets to coalesce */
312 #define TF_NOOPT 0x000008 /* don't use tcp options */
313 #define TF_SENTFIN 0x000010 /* have sent FIN */
314 #define TF_REQ_SCALE 0x000020 /* have/will request window scaling */
315 #define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */
316 #define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */
317 #define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */
318 #define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */
319 #define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */
320 #define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */
321 #define TF_NOPUSH 0x001000 /* don't push */
322 #define TF_PREVVALID 0x002000 /* saved values for bad rxmit valid */
323 #define TF_MORETOCOME 0x010000 /* More data to be appended to sock */
324 #define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */
325 #define TF_LASTIDLE 0x040000 /* connection was previously idle */
326 #define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */
327 #define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */
328 #define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */
329 #define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */
330 #define TF_FORCEDATA 0x800000 /* force out a byte */
331 #define TF_TSO 0x1000000 /* TSO enabled on this connection */
332 #define TF_TOE 0x2000000 /* this connection is offloaded */
333 #define TF_ECN_PERMIT 0x4000000 /* connection ECN-ready */
334 #define TF_ECN_SND_CWR 0x8000000 /* ECN CWR in queue */
335 #define TF_ECN_SND_ECE 0x10000000 /* ECN ECE in queue */
336 #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */
337 #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */
338 #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */
340 #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY)
341 #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY
342 #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY
344 #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY)
345 #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY
346 #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY
348 #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
349 #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
350 #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
352 #if defined(_KERNEL) && !defined(TCP_RFC7413)
353 #define IS_FASTOPEN(t_flags) (false)
355 #define IS_FASTOPEN(t_flags) (t_flags & TF_FASTOPEN)
358 #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una)
361 * Flags for the t_oobflags field.
363 #define TCPOOB_HAVEDATA 0x01
364 #define TCPOOB_HADDATA 0x02
367 * Flags for the extended TCP flags field, t_flags2
369 #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */
370 #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */
371 #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */
372 #define TF2_LOG_AUTO 0x00000008 /* Session is auto-logging. */
373 #define TF2_DROP_AF_DATA 0x00000010 /* Drop after all data ack'd */
376 * Structure to hold TCP options that are only used during segment
377 * processing (in tcp_input), but not held in the tcpcb.
378 * It's basically used to reduce the number of parameters
379 * to tcp_dooptions and tcp_addoptions.
380 * The binary order of the to_flags is relevant for packing of the
381 * options in tcp_addoptions.
384 u_int32_t to_flags; /* which options are present */
385 #define TOF_MSS 0x0001 /* maximum segment size */
386 #define TOF_SCALE 0x0002 /* window scaling */
387 #define TOF_SACKPERM 0x0004 /* SACK permitted */
388 #define TOF_TS 0x0010 /* timestamp */
389 #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */
390 #define TOF_SACK 0x0080 /* Peer sent SACK option */
391 #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */
392 #define TOF_MAXOPT 0x0200
393 u_int32_t to_tsval; /* new timestamp */
394 u_int32_t to_tsecr; /* reflected timestamp */
395 u_char *to_sacks; /* pointer to the first SACK blocks */
396 u_char *to_signature; /* pointer to the TCP-MD5 signature */
397 u_int8_t *to_tfo_cookie; /* pointer to the TFO cookie */
398 u_int16_t to_mss; /* maximum segment size */
399 u_int8_t to_wscale; /* window scaling */
400 u_int8_t to_nsacks; /* number of SACK blocks */
401 u_int8_t to_tfo_len; /* TFO cookie length */
402 u_int32_t to_spare; /* UTO */
406 * Flags for tcp_dooptions.
408 #define TO_SYN 0x01 /* parse SYN-only options */
410 struct hc_metrics_lite { /* must stay in sync with hc_metrics */
411 uint32_t rmx_mtu; /* MTU for this path */
412 uint32_t rmx_ssthresh; /* outbound gateway buffer limit */
413 uint32_t rmx_rtt; /* estimated round trip time */
414 uint32_t rmx_rttvar; /* estimated rtt variance */
415 uint32_t rmx_cwnd; /* congestion window */
416 uint32_t rmx_sendpipe; /* outbound delay-bandwidth product */
417 uint32_t rmx_recvpipe; /* inbound delay-bandwidth product */
421 * Used by tcp_maxmtu() to communicate interface specific features
422 * and limits at the time of connection setup.
427 u_int tsomaxsegcount;
431 #ifndef _NETINET_IN_PCB_H_
433 #endif /* _NETINET_IN_PCB_H_ */
436 struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */
441 u_short last_win; /* cached window value */
442 short tw_so_options; /* copy of so_options */
443 struct ucred *tw_cred; /* user credentials */
445 u_int32_t ts_offset; /* our timestamp offset */
448 TAILQ_ENTRY(tcptw) tw_2msl;
449 void *tw_pspare; /* TCP_SIGNATURE */
450 u_int *tw_spare; /* TCP_SIGNATURE */
453 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
454 #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb)
455 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
458 * The smoothed round-trip time and estimated variance
459 * are stored as fixed point numbers scaled by the values below.
460 * For convenience, these scales are also used in smoothing the average
461 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
462 * With these scales, srtt has 3 bits to the right of the binary point,
463 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
464 * binary point, and is smoothed with an ALPHA of 0.75.
466 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
467 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
468 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
469 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
470 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
473 * The initial retransmission should happen at rtt + 4 * rttvar.
474 * Because of the way we do the smoothing, srtt and rttvar
475 * will each average +1/2 tick of bias. When we compute
476 * the retransmit timer, we want 1/2 tick of rounding and
477 * 1 extra tick because of +-1/2 tick uncertainty in the
478 * firing of the timer. The bias will give us exactly the
479 * 1.5 tick we need. But, because the bias is
480 * statistical, we have to test that we don't drop below
481 * the minimum feasible timer (which is 2 ticks).
482 * This version of the macro adapted from a paper by Lawrence
483 * Brakmo and Larry Peterson which outlines a problem caused
484 * by insufficient precision in the original implementation,
485 * which results in inappropriately large RTO values for very
488 #define TCP_REXMTVAL(tp) \
489 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
490 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
494 * Many of these should be kept per connection,
495 * but that's inconvenient at the moment.
498 uint64_t tcps_connattempt; /* connections initiated */
499 uint64_t tcps_accepts; /* connections accepted */
500 uint64_t tcps_connects; /* connections established */
501 uint64_t tcps_drops; /* connections dropped */
502 uint64_t tcps_conndrops; /* embryonic connections dropped */
503 uint64_t tcps_minmssdrops; /* average minmss too low drops */
504 uint64_t tcps_closed; /* conn. closed (includes drops) */
505 uint64_t tcps_segstimed; /* segs where we tried to get rtt */
506 uint64_t tcps_rttupdated; /* times we succeeded */
507 uint64_t tcps_delack; /* delayed acks sent */
508 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
509 uint64_t tcps_rexmttimeo; /* retransmit timeouts */
510 uint64_t tcps_persisttimeo; /* persist timeouts */
511 uint64_t tcps_keeptimeo; /* keepalive timeouts */
512 uint64_t tcps_keepprobe; /* keepalive probes sent */
513 uint64_t tcps_keepdrops; /* connections dropped in keepalive */
515 uint64_t tcps_sndtotal; /* total packets sent */
516 uint64_t tcps_sndpack; /* data packets sent */
517 uint64_t tcps_sndbyte; /* data bytes sent */
518 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */
519 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */
520 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
521 uint64_t tcps_sndacks; /* ack-only packets sent */
522 uint64_t tcps_sndprobe; /* window probes sent */
523 uint64_t tcps_sndurg; /* packets sent with URG only */
524 uint64_t tcps_sndwinup; /* window update-only packets sent */
525 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
527 uint64_t tcps_rcvtotal; /* total packets received */
528 uint64_t tcps_rcvpack; /* packets received in sequence */
529 uint64_t tcps_rcvbyte; /* bytes received in sequence */
530 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */
531 uint64_t tcps_rcvbadoff; /* packets received with bad offset */
532 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */
533 uint64_t tcps_rcvshort; /* packets received too short */
534 uint64_t tcps_rcvduppack; /* duplicate-only packets received */
535 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */
536 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */
537 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
538 uint64_t tcps_rcvoopack; /* out-of-order packets received */
539 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */
540 uint64_t tcps_rcvpackafterwin; /* packets with data after window */
541 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
542 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */
543 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */
544 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */
545 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
546 uint64_t tcps_rcvackpack; /* rcvd ack packets */
547 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
548 uint64_t tcps_rcvwinupd; /* rcvd window update packets */
549 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */
550 uint64_t tcps_predack; /* times hdr predict ok for acks */
551 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */
552 uint64_t tcps_pcbcachemiss;
553 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */
554 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */
555 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */
556 uint64_t tcps_usedrtt; /* times RTT initialized from route */
557 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
558 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
559 uint64_t tcps_persistdrop; /* timeout in persist state */
560 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
561 uint64_t tcps_mturesent; /* resends due to MTU discovery */
562 uint64_t tcps_listendrop; /* listen queue overflows */
563 uint64_t tcps_badrst; /* ignored RSTs in the window */
565 uint64_t tcps_sc_added; /* entry added to syncache */
566 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
567 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */
568 uint64_t tcps_sc_dropped; /* could not reply to packet */
569 uint64_t tcps_sc_completed; /* successful extraction of entry */
570 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
571 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
572 uint64_t tcps_sc_reset; /* RST removed entry from syncache */
573 uint64_t tcps_sc_stale; /* timed out or listen socket gone */
574 uint64_t tcps_sc_aborted; /* syncache entry aborted */
575 uint64_t tcps_sc_badack; /* removed due to bad ACK */
576 uint64_t tcps_sc_unreach; /* ICMP unreachable received */
577 uint64_t tcps_sc_zonefail; /* zalloc() failed */
578 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */
579 uint64_t tcps_sc_recvcookie; /* SYN cookie received */
581 uint64_t tcps_hc_added; /* entry added to hostcache */
582 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
584 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */
586 /* SACK related stats */
587 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
588 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */
589 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
590 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
591 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
592 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */
594 /* ECN related stats */
595 uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */
596 uint64_t tcps_ecn_ect0; /* ECN Capable Transport */
597 uint64_t tcps_ecn_ect1; /* ECN Capable Transport */
598 uint64_t tcps_ecn_shs; /* ECN successful handshakes */
599 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */
601 /* TCP_SIGNATURE related stats */
602 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */
603 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */
604 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */
605 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */
606 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */
608 /* Path MTU Discovery Black Hole Detection related stats */
609 uint64_t tcps_pmtud_blackhole_activated; /* Black Hole Count */
610 uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
611 uint64_t tcps_pmtud_blackhole_failed; /* Black Hole Failure Count */
613 uint64_t _pad[12]; /* 6 UTO, 6 TBD */
616 #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */
619 #define TI_UNLOCKED 1
621 #include <sys/counter.h>
623 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */
625 * In-kernel consumers can use these accessor macros directly to update
628 #define TCPSTAT_ADD(name, val) \
629 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
630 #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1)
633 * Kernel module consumers must use this accessor macro.
635 void kmod_tcpstat_inc(int statnum);
636 #define KMOD_TCPSTAT_INC(name) \
637 kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
640 * Running TCP connection count by state.
642 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
643 #define V_tcps_states VNET(tcps_states)
644 #define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1)
645 #define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1)
648 * TCP specific helper hook point identifiers.
650 #define HHOOK_TCP_EST_IN 0
651 #define HHOOK_TCP_EST_OUT 1
652 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT
654 struct tcp_hhook_data {
663 void hhook_run_tcp_est_out(struct tcpcb *tp,
664 struct tcphdr *th, struct tcpopt *to,
665 uint32_t len, int tso);
670 * TCB structure exported to user-land via sysctl(3).
672 * Fields prefixed with "xt_" are unique to the export structure, and fields
673 * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
676 * (s) - used by userland utilities in src
677 * (p) - used by utilities in ports
678 * (3) - is known to be used by third party software not in ports
679 * (n) - no known usage
681 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
682 * included. Not all of our clients do.
684 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
686 ksize_t xt_len; /* length of this structure */
687 struct xinpcb xt_inp;
688 char xt_stack[TCP_FUNCTION_NAME_LEN_MAX]; /* (s) */
689 char xt_logid[TCP_LOG_ID_LEN]; /* (s) */
691 int32_t t_state; /* (s,p) */
692 uint32_t t_flags; /* (s,p) */
693 int32_t t_sndzerowin; /* (s) */
694 int32_t t_sndrexmitpack; /* (s) */
695 int32_t t_rcvoopack; /* (s) */
696 int32_t t_rcvtime; /* (s) */
697 int32_t tt_rexmt; /* (s) */
698 int32_t tt_persist; /* (s) */
699 int32_t tt_keep; /* (s) */
700 int32_t tt_2msl; /* (s) */
701 int32_t tt_delack; /* (s) */
702 int32_t t_logstate; /* (3) */
707 void tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
712 * TCP function information (name-to-id mapping, aliases, and refcnt)
713 * exported to user-land via sysctl(3).
715 struct tcp_function_info {
718 char tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
719 char tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
723 * Identifiers for TCP sysctl nodes
725 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
726 #define TCPCTL_MSSDFLT 3 /* MSS default */
727 #define TCPCTL_STATS 4 /* statistics */
728 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
729 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
730 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
731 #define TCPCTL_SENDSPACE 8 /* send buffer space */
732 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
733 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
734 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
735 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
736 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
737 #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */
738 #define TCPCTL_DROP 15 /* drop tcp connection */
739 #define TCPCTL_STATES 16 /* connection counts by TCP state */
743 SYSCTL_DECL(_net_inet_tcp);
744 SYSCTL_DECL(_net_inet_tcp_sack);
745 MALLOC_DECLARE(M_TCPLOG);
748 extern int tcp_log_in_vain;
751 * Global TCP tunables shared between different stacks.
752 * Please keep the list sorted.
754 VNET_DECLARE(int, drop_synfin);
755 VNET_DECLARE(int, path_mtu_discovery);
756 VNET_DECLARE(int, tcp_abc_l_var);
757 VNET_DECLARE(int, tcp_autorcvbuf_max);
758 VNET_DECLARE(int, tcp_autosndbuf_inc);
759 VNET_DECLARE(int, tcp_autosndbuf_max);
760 VNET_DECLARE(int, tcp_delack_enabled);
761 VNET_DECLARE(int, tcp_do_autorcvbuf);
762 VNET_DECLARE(int, tcp_do_autosndbuf);
763 VNET_DECLARE(int, tcp_do_ecn);
764 VNET_DECLARE(int, tcp_do_rfc1323);
765 VNET_DECLARE(int, tcp_do_rfc3042);
766 VNET_DECLARE(int, tcp_do_rfc3390);
767 VNET_DECLARE(int, tcp_do_rfc3465);
768 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
769 VNET_DECLARE(int, tcp_do_sack);
770 VNET_DECLARE(int, tcp_do_tso);
771 VNET_DECLARE(int, tcp_ecn_maxretries);
772 VNET_DECLARE(int, tcp_initcwnd_segments);
773 VNET_DECLARE(int, tcp_insecure_rst);
774 VNET_DECLARE(int, tcp_insecure_syn);
775 VNET_DECLARE(int, tcp_minmss);
776 VNET_DECLARE(int, tcp_mssdflt);
777 VNET_DECLARE(int, tcp_recvspace);
778 VNET_DECLARE(int, tcp_sack_globalholes);
779 VNET_DECLARE(int, tcp_sack_globalmaxholes);
780 VNET_DECLARE(int, tcp_sack_maxholes);
781 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
782 VNET_DECLARE(int, tcp_sendspace);
783 VNET_DECLARE(struct inpcbhead, tcb);
784 VNET_DECLARE(struct inpcbinfo, tcbinfo);
786 #define V_drop_synfin VNET(drop_synfin)
787 #define V_path_mtu_discovery VNET(path_mtu_discovery)
788 #define V_tcb VNET(tcb)
789 #define V_tcbinfo VNET(tcbinfo)
790 #define V_tcp_abc_l_var VNET(tcp_abc_l_var)
791 #define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max)
792 #define V_tcp_autosndbuf_inc VNET(tcp_autosndbuf_inc)
793 #define V_tcp_autosndbuf_max VNET(tcp_autosndbuf_max)
794 #define V_tcp_delack_enabled VNET(tcp_delack_enabled)
795 #define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf)
796 #define V_tcp_do_autosndbuf VNET(tcp_do_autosndbuf)
797 #define V_tcp_do_ecn VNET(tcp_do_ecn)
798 #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323)
799 #define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042)
800 #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390)
801 #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465)
802 #define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe)
803 #define V_tcp_do_sack VNET(tcp_do_sack)
804 #define V_tcp_do_tso VNET(tcp_do_tso)
805 #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries)
806 #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments)
807 #define V_tcp_insecure_rst VNET(tcp_insecure_rst)
808 #define V_tcp_insecure_syn VNET(tcp_insecure_syn)
809 #define V_tcp_minmss VNET(tcp_minmss)
810 #define V_tcp_mssdflt VNET(tcp_mssdflt)
811 #define V_tcp_recvspace VNET(tcp_recvspace)
812 #define V_tcp_sack_globalholes VNET(tcp_sack_globalholes)
813 #define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes)
814 #define V_tcp_sack_maxholes VNET(tcp_sack_maxholes)
815 #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail)
816 #define V_tcp_sendspace VNET(tcp_sendspace)
817 #define V_tcp_udp_tunneling_overhead VNET(tcp_udp_tunneling_overhead)
818 #define V_tcp_udp_tunneling_port VNET(tcp_udp_tunneling_port)
822 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
823 #define V_tcp_hhh VNET(tcp_hhh)
826 int tcp_addoptions(struct tcpopt *, u_char *);
827 int tcp_ccalgounload(struct cc_algo *unload_algo);
829 tcp_close(struct tcpcb *);
830 void tcp_discardcb(struct tcpcb *);
831 void tcp_twstart(struct tcpcb *);
832 void tcp_twclose(struct tcptw *, int);
833 void tcp_ctlinput(int, struct sockaddr *, void *);
834 int tcp_ctloutput(struct socket *, struct sockopt *);
836 tcp_drop(struct tcpcb *, int);
837 void tcp_drain(void);
839 void tcp_fini(void *);
840 char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
842 char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
844 int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *, struct mbuf *);
845 void tcp_reass_global_init(void);
846 void tcp_reass_flush(struct tcpcb *);
847 void tcp_dooptions(struct tcpopt *, u_char *, int, int);
848 void tcp_dropwithreset(struct mbuf *, struct tcphdr *,
849 struct tcpcb *, int, int);
850 void tcp_pulloutofband(struct socket *,
851 struct tcphdr *, struct mbuf *, int);
852 void tcp_xmit_timer(struct tcpcb *, int);
853 void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
854 void cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
855 uint16_t nsegs, uint16_t type);
856 void cc_conn_init(struct tcpcb *tp);
857 void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
858 void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
860 void hhook_run_tcp_est_in(struct tcpcb *tp,
861 struct tcphdr *th, struct tcpopt *to);
864 int tcp_input(struct mbuf **, int *, int);
865 int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
866 struct tcpcb *, int);
867 void tcp_do_segment(struct mbuf *, struct tcphdr *,
868 struct socket *, struct tcpcb *, int, int, uint8_t);
870 int register_tcp_functions(struct tcp_function_block *blk, int wait);
871 int register_tcp_functions_as_names(struct tcp_function_block *blk,
872 int wait, const char *names[], int *num_names);
873 int register_tcp_functions_as_name(struct tcp_function_block *blk,
874 const char *name, int wait);
875 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
877 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
878 void tcp_switch_back_to_default(struct tcpcb *tp);
879 struct tcp_function_block *
880 find_and_ref_tcp_fb(struct tcp_function_block *fs);
881 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
883 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
884 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
885 u_int tcp_maxseg(const struct tcpcb *);
886 void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
888 void tcp_mss(struct tcpcb *, int);
889 int tcp_mssopt(struct in_conninfo *);
891 tcp_drop_syn_sent(struct inpcb *, int);
893 tcp_newtcpcb(struct inpcb *);
894 int tcp_output(struct tcpcb *);
895 void tcp_state_change(struct tcpcb *, int);
896 void tcp_respond(struct tcpcb *, void *,
897 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
898 void tcp_tw_init(void);
900 void tcp_tw_destroy(void);
902 void tcp_tw_zone_change(void);
903 int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
905 void tcp_setpersist(struct tcpcb *);
906 void tcp_slowtimo(void);
908 tcpip_maketemplate(struct inpcb *);
909 void tcpip_fillheaders(struct inpcb *, void *, void *);
910 void tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
911 int tcp_timer_suspend(struct tcpcb *, uint32_t);
912 void tcp_timers_unsuspend(struct tcpcb *, uint32_t);
913 int tcp_timer_active(struct tcpcb *, uint32_t);
914 void tcp_timer_stop(struct tcpcb *, uint32_t);
915 void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
916 int inp_to_cpuid(struct inpcb *inp);
918 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
920 void tcp_hc_init(void);
922 void tcp_hc_destroy(void);
924 void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
925 uint32_t tcp_hc_getmtu(struct in_conninfo *);
926 void tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
927 void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
929 extern struct pr_usrreqs tcp_usrreqs;
931 uint32_t tcp_new_ts_offset(struct in_conninfo *);
932 tcp_seq tcp_new_isn(struct in_conninfo *);
934 int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
935 void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
936 void tcp_clean_sackreport(struct tcpcb *tp);
937 void tcp_sack_adjust(struct tcpcb *tp);
938 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
939 void tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
940 void tcp_free_sackholes(struct tcpcb *tp);
941 int tcp_newreno(struct tcpcb *, struct tcphdr *);
942 int tcp_compute_pipe(struct tcpcb *);
943 uint32_t tcp_compute_initwnd(uint32_t);
944 void tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
946 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
947 int32_t seglimit, int32_t segsize, struct sockbuf *sb);
951 tcp_fields_to_host(struct tcphdr *th)
954 th->th_seq = ntohl(th->th_seq);
955 th->th_ack = ntohl(th->th_ack);
956 th->th_win = ntohs(th->th_win);
957 th->th_urp = ntohs(th->th_urp);
961 tcp_fields_to_net(struct tcphdr *th)
964 th->th_seq = htonl(th->th_seq);
965 th->th_ack = htonl(th->th_ack);
966 th->th_win = htons(th->th_win);
967 th->th_urp = htons(th->th_urp);
971 #endif /* _NETINET_TCP_VAR_H_ */