xref: /freebsd/sys/netinet/tcp_var.h (revision 74ca7bf1d4c7173d5575ba168bc4b5f6d181ff5a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1993, 1994, 1995
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
32  * $FreeBSD$
33  */
34 
35 #ifndef _NETINET_TCP_VAR_H_
36 #define _NETINET_TCP_VAR_H_
37 
38 #include <netinet/tcp.h>
39 #include <netinet/tcp_fsm.h>
40 
41 #ifdef _KERNEL
42 #include <net/vnet.h>
43 #include <sys/mbuf.h>
44 #endif
45 
46 #if defined(_KERNEL) || defined(_WANT_TCPCB)
47 /* TCP segment queue entry */
48 struct tseg_qent {
49 	LIST_ENTRY(tseg_qent) tqe_q;
50 	int	tqe_len;		/* TCP segment data length */
51 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
52 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
53 };
54 LIST_HEAD(tsegqe_head, tseg_qent);
55 
56 struct sackblk {
57 	tcp_seq start;		/* start seq no. of sack block */
58 	tcp_seq end;		/* end seq no. */
59 };
60 
61 struct sackhole {
62 	tcp_seq start;		/* start seq no. of hole */
63 	tcp_seq end;		/* end seq no. */
64 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
65 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
66 };
67 
68 struct sackhint {
69 	struct sackhole	*nexthole;
70 	int		sack_bytes_rexmit;
71 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
72 
73 	int		ispare;		/* explicit pad for 64bit alignment */
74 	int             sacked_bytes;	/*
75 					 * Total sacked bytes reported by the
76 					 * receiver via sack option
77 					 */
78 	uint32_t	_pad1[1];	/* TBD */
79 	uint64_t	_pad[1];	/* TBD */
80 };
81 
82 STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
83 
84 /*
85  * Tcp control block, one per tcp; fields:
86  * Organized for 16 byte cacheline efficiency.
87  */
88 struct tcpcb {
89 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
90 	int	t_segqlen;		/* segment reassembly queue length */
91 	int	t_dupacks;		/* consecutive dup acks recd */
92 
93 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
94 
95 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
96 	int	t_state;		/* state of this connection */
97 	u_int	t_flags;
98 
99 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
100 
101 	tcp_seq	snd_una;		/* sent but unacknowledged */
102 	tcp_seq	snd_max;		/* highest sequence number sent;
103 					 * used to recognize retransmits
104 					 */
105 	tcp_seq	snd_nxt;		/* send next */
106 	tcp_seq	snd_up;			/* send urgent pointer */
107 
108 	tcp_seq	snd_wl1;		/* window update seg seq number */
109 	tcp_seq	snd_wl2;		/* window update seg ack number */
110 	tcp_seq	iss;			/* initial send sequence number */
111 	tcp_seq	irs;			/* initial receive sequence number */
112 
113 	tcp_seq	rcv_nxt;		/* receive next */
114 	tcp_seq	rcv_adv;		/* advertised window */
115 	uint32_t  rcv_wnd;		/* receive window */
116 	tcp_seq	rcv_up;			/* receive urgent pointer */
117 
118 	uint32_t  snd_wnd;		/* send window */
119 	uint32_t  snd_cwnd;		/* congestion-controlled window */
120 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
121 					 * for slow start exponential to
122 					 * linear switch
123 					 */
124 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
125 
126 	u_int	t_rcvtime;		/* inactivity time */
127 	u_int	t_starttime;		/* time connection was established */
128 	u_int	t_rtttime;		/* RTT measurement start time */
129 	tcp_seq	t_rtseq;		/* sequence number being timed */
130 
131 	int	t_rxtcur;		/* current retransmit value (ticks) */
132 	u_int	t_maxseg;		/* maximum segment size */
133 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
134 	int	t_srtt;			/* smoothed round-trip time */
135 	int	t_rttvar;		/* variance in round-trip time */
136 
137 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
138 	u_int	t_rttmin;		/* minimum rtt allowed */
139 	u_int	t_rttbest;		/* best rtt we've seen */
140 	u_long	t_rttupdated;		/* number of times rtt sampled */
141 	uint32_t  max_sndwnd;		/* largest window peer has offered */
142 
143 	int	t_softerror;		/* possible error not yet reported */
144 /* out-of-band data */
145 	char	t_oobflags;		/* have some */
146 	char	t_iobc;			/* input character */
147 /* RFC 1323 variables */
148 	u_char	snd_scale;		/* window scaling for send window */
149 	u_char	rcv_scale;		/* window scaling for recv window */
150 	u_char	request_r_scale;	/* pending window scaling */
151 	u_int32_t  ts_recent;		/* timestamp echo data */
152 	u_int	ts_recent_age;		/* when last updated */
153 	u_int32_t  ts_offset;		/* our timestamp offset */
154 
155 	tcp_seq	last_ack_sent;
156 /* experimental */
157 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
158 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
159 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
160 	int	t_sndzerowin;		/* zero-window updates sent */
161 	u_int	t_badrxtwin;		/* window for retransmit recovery */
162 	u_char	snd_limited;		/* segments limited transmitted */
163 /* SACK related state */
164 	int	snd_numholes;		/* number of holes seen by sender */
165 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
166 					/* SACK scoreboard (sorted) */
167 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
168 	int	rcv_numsacks;		/* # distinct sack blks present */
169 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
170 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
171 					   episode starts at this seq number */
172 	struct sackhint	sackhint;	/* SACK scoreboard hint */
173 	int	t_rttlow;		/* smallest observerved RTT */
174 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
175 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
176 	struct toedev	*tod;		/* toedev handling this connection */
177 	int	t_sndrexmitpack;	/* retransmit packets sent */
178 	int	t_rcvoopack;		/* out-of-order packets received */
179 	void	*t_toe;			/* TOE pcb pointer */
180 	int	t_bytes_acked;		/* # bytes acked during current RTT */
181 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
182 	struct cc_var	*ccv;		/* congestion control specific vars */
183 	struct osd	*osd;		/* storage for Khelp module data */
184 
185 	u_int	t_keepinit;		/* time to establish connection */
186 	u_int	t_keepidle;		/* time before keepalive probes begin */
187 	u_int	t_keepintvl;		/* interval between keepalives */
188 	u_int	t_keepcnt;		/* number of keepalives before close */
189 
190 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
191 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
192 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
193 	u_int	t_flags2;		/* More tcpcb flags storage */
194 	int	t_logstate;		/* State of "black box" logging */
195 	struct tcp_log_stailq t_logs;	/* Log buffer */
196 	int	t_lognum;		/* Number of log entries */
197 	uint32_t t_logsn;		/* Log "serial number" */
198 	struct tcp_log_id_node *t_lin;
199 	struct tcp_log_id_bucket *t_lib;
200 	const char *t_output_caller;	/* Function that called tcp_output */
201 	struct tcp_function_block *t_fb;/* TCP function call block */
202 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
203 	uint8_t t_tfo_client_cookie_len; /* TCP Fast Open client cookie length */
204 	unsigned int *t_tfo_pending;	/* TCP Fast Open server pending counter */
205 	union {
206 		uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
207 		uint64_t server;
208 	} t_tfo_cookie;			/* TCP Fast Open cookie to send */
209 #ifdef TCPPCAP
210 	struct mbufq t_inpkts;		/* List of saved input packets. */
211 	struct mbufq t_outpkts;		/* List of saved output packets. */
212 #endif
213 };
214 #endif	/* _KERNEL || _WANT_TCPCB */
215 
216 #ifdef _KERNEL
217 struct tcptemp {
218 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
219 	struct	tcphdr tt_t;
220 };
221 
222 /*
223  * TODO: We yet need to brave plowing in
224  * to tcp_input() and the pru_usrreq() block.
225  * Right now these go to the old standards which
226  * are somewhat ok, but in the long term may
227  * need to be changed. If we do tackle tcp_input()
228  * then we need to get rid of the tcp_do_segment()
229  * function below.
230  */
231 /* Flags for tcp functions */
232 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
233 
234 /*
235  * If defining the optional tcp_timers, in the
236  * tfb_tcp_timer_stop call you must use the
237  * callout_async_drain() function with the
238  * tcp_timer_discard callback. You should check
239  * the return of callout_async_drain() and if 0
240  * increment tt_draincnt. Since the timer sub-system
241  * does not know your callbacks you must provide a
242  * stop_all function that loops through and calls
243  * tcp_timer_stop() with each of your defined timers.
244  * Adding a tfb_tcp_handoff_ok function allows the socket
245  * option to change stacks to query you even if the
246  * connection is in a later stage. You return 0 to
247  * say you can take over and run your stack, you return
248  * non-zero (an error number) to say no you can't.
249  * If the function is undefined you can only change
250  * in the early states (before connect or listen).
251  * tfb_tcp_fb_fini is changed to add a flag to tell
252  * the old stack if the tcb is being destroyed or
253  * not. A one in the flag means the TCB is being
254  * destroyed, a zero indicates its transitioning to
255  * another stack (via socket option).
256  */
257 struct tcp_function_block {
258 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
259 	int	(*tfb_tcp_output)(struct tcpcb *);
260 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
261 			    struct socket *, struct tcpcb *,
262 			    int, int, uint8_t,
263 			    int);
264 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
265 			    struct inpcb *inp, struct tcpcb *tp);
266 	/* Optional memory allocation/free routine */
267 	void	(*tfb_tcp_fb_init)(struct tcpcb *);
268 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
269 	/* Optional timers, must define all if you define one */
270 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
271 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
272 			    uint32_t, u_int);
273 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
274 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
275 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
276 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
277 	volatile uint32_t tfb_refcnt;
278 	uint32_t  tfb_flags;
279 	uint8_t	tfb_id;
280 };
281 
282 struct tcp_function {
283 	TAILQ_ENTRY(tcp_function)	tf_next;
284 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
285 	struct tcp_function_block	*tf_fb;
286 };
287 
288 TAILQ_HEAD(tcp_funchead, tcp_function);
289 #endif	/* _KERNEL */
290 
291 /*
292  * Flags and utility macros for the t_flags field.
293  */
294 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
295 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
296 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
297 #define	TF_NOOPT	0x000008	/* don't use tcp options */
298 #define	TF_SENTFIN	0x000010	/* have sent FIN */
299 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
300 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
301 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
302 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
303 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
304 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
305 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
306 #define	TF_NOPUSH	0x001000	/* don't push */
307 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
308 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
309 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
310 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
311 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
312 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
313 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
314 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
315 #define	TF_FORCEDATA	0x800000	/* force out a byte */
316 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
317 #define	TF_TOE		0x2000000	/* this connection is offloaded */
318 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
319 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
320 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
321 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
322 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
323 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
324 
325 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
326 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
327 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
328 
329 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
330 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
331 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
332 
333 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
334 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
335 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
336 
337 #if defined(_KERNEL) && !defined(TCP_RFC7413)
338 #define	IS_FASTOPEN(t_flags)		(false)
339 #else
340 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
341 #endif
342 
343 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
344 
345 /*
346  * Flags for the t_oobflags field.
347  */
348 #define	TCPOOB_HAVEDATA	0x01
349 #define	TCPOOB_HADDATA	0x02
350 
351 /*
352  * Flags for the extended TCP flags field, t_flags2
353  */
354 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
355 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
356 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
357 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
358 
359 /*
360  * Structure to hold TCP options that are only used during segment
361  * processing (in tcp_input), but not held in the tcpcb.
362  * It's basically used to reduce the number of parameters
363  * to tcp_dooptions and tcp_addoptions.
364  * The binary order of the to_flags is relevant for packing of the
365  * options in tcp_addoptions.
366  */
367 struct tcpopt {
368 	u_int32_t	to_flags;	/* which options are present */
369 #define	TOF_MSS		0x0001		/* maximum segment size */
370 #define	TOF_SCALE	0x0002		/* window scaling */
371 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
372 #define	TOF_TS		0x0010		/* timestamp */
373 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
374 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
375 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
376 #define	TOF_MAXOPT	0x0200
377 	u_int32_t	to_tsval;	/* new timestamp */
378 	u_int32_t	to_tsecr;	/* reflected timestamp */
379 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
380 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
381 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
382 	u_int16_t	to_mss;		/* maximum segment size */
383 	u_int8_t	to_wscale;	/* window scaling */
384 	u_int8_t	to_nsacks;	/* number of SACK blocks */
385 	u_int8_t	to_tfo_len;	/* TFO cookie length */
386 	u_int32_t	to_spare;	/* UTO */
387 };
388 
389 /*
390  * Flags for tcp_dooptions.
391  */
392 #define	TO_SYN		0x01		/* parse SYN-only options */
393 
394 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
395 	uint32_t	rmx_mtu;	/* MTU for this path */
396 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
397 	uint32_t	rmx_rtt;	/* estimated round trip time */
398 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
399 	uint32_t	rmx_cwnd;	/* congestion window */
400 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
401 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
402 };
403 
404 /*
405  * Used by tcp_maxmtu() to communicate interface specific features
406  * and limits at the time of connection setup.
407  */
408 struct tcp_ifcap {
409 	int	ifcap;
410 	u_int	tsomax;
411 	u_int	tsomaxsegcount;
412 	u_int	tsomaxsegsize;
413 };
414 
415 #ifndef _NETINET_IN_PCB_H_
416 struct in_conninfo;
417 #endif /* _NETINET_IN_PCB_H_ */
418 
419 struct tcptw {
420 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
421 	tcp_seq		snd_nxt;
422 	tcp_seq		rcv_nxt;
423 	tcp_seq		iss;
424 	tcp_seq		irs;
425 	u_short		last_win;	/* cached window value */
426 	short		tw_so_options;	/* copy of so_options */
427 	struct ucred	*tw_cred;	/* user credentials */
428 	u_int32_t	t_recent;
429 	u_int32_t	ts_offset;	/* our timestamp offset */
430 	u_int		t_starttime;
431 	int		tw_time;
432 	TAILQ_ENTRY(tcptw) tw_2msl;
433 	void		*tw_pspare;	/* TCP_SIGNATURE */
434 	u_int		*tw_spare;	/* TCP_SIGNATURE */
435 };
436 
437 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
438 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
439 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
440 
441 /*
442  * The smoothed round-trip time and estimated variance
443  * are stored as fixed point numbers scaled by the values below.
444  * For convenience, these scales are also used in smoothing the average
445  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
446  * With these scales, srtt has 3 bits to the right of the binary point,
447  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
448  * binary point, and is smoothed with an ALPHA of 0.75.
449  */
450 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
451 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
452 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
453 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
454 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
455 
456 /*
457  * The initial retransmission should happen at rtt + 4 * rttvar.
458  * Because of the way we do the smoothing, srtt and rttvar
459  * will each average +1/2 tick of bias.  When we compute
460  * the retransmit timer, we want 1/2 tick of rounding and
461  * 1 extra tick because of +-1/2 tick uncertainty in the
462  * firing of the timer.  The bias will give us exactly the
463  * 1.5 tick we need.  But, because the bias is
464  * statistical, we have to test that we don't drop below
465  * the minimum feasible timer (which is 2 ticks).
466  * This version of the macro adapted from a paper by Lawrence
467  * Brakmo and Larry Peterson which outlines a problem caused
468  * by insufficient precision in the original implementation,
469  * which results in inappropriately large RTO values for very
470  * fast networks.
471  */
472 #define	TCP_REXMTVAL(tp) \
473 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
474 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
475 
476 /*
477  * TCP statistics.
478  * Many of these should be kept per connection,
479  * but that's inconvenient at the moment.
480  */
481 struct	tcpstat {
482 	uint64_t tcps_connattempt;	/* connections initiated */
483 	uint64_t tcps_accepts;		/* connections accepted */
484 	uint64_t tcps_connects;		/* connections established */
485 	uint64_t tcps_drops;		/* connections dropped */
486 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
487 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
488 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
489 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
490 	uint64_t tcps_rttupdated;	/* times we succeeded */
491 	uint64_t tcps_delack;		/* delayed acks sent */
492 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
493 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
494 	uint64_t tcps_persisttimeo;	/* persist timeouts */
495 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
496 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
497 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
498 
499 	uint64_t tcps_sndtotal;		/* total packets sent */
500 	uint64_t tcps_sndpack;		/* data packets sent */
501 	uint64_t tcps_sndbyte;		/* data bytes sent */
502 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
503 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
504 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
505 	uint64_t tcps_sndacks;		/* ack-only packets sent */
506 	uint64_t tcps_sndprobe;		/* window probes sent */
507 	uint64_t tcps_sndurg;		/* packets sent with URG only */
508 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
509 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
510 
511 	uint64_t tcps_rcvtotal;		/* total packets received */
512 	uint64_t tcps_rcvpack;		/* packets received in sequence */
513 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
514 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
515 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
516 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
517 	uint64_t tcps_rcvshort;		/* packets received too short */
518 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
519 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
520 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
521 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
522 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
523 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
524 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
525 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
526 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
527 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
528 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
529 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
530 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
531 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
532 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
533 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
534 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
535 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
536 	uint64_t tcps_pcbcachemiss;
537 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
538 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
539 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
540 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
541 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
542 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
543 	uint64_t tcps_persistdrop;	/* timeout in persist state */
544 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
545 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
546 	uint64_t tcps_listendrop;	/* listen queue overflows */
547 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
548 
549 	uint64_t tcps_sc_added;		/* entry added to syncache */
550 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
551 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
552 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
553 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
554 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
555 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
556 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
557 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
558 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
559 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
560 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
561 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
562 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
563 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
564 
565 	uint64_t tcps_hc_added;		/* entry added to hostcache */
566 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
567 
568 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
569 
570 	/* SACK related stats */
571 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
572 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
573 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
574 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
575 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
576 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
577 
578 	/* ECN related stats */
579 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
580 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
581 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
582 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
583 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
584 
585 	/* TCP_SIGNATURE related stats */
586 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
587 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
588 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
589 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
590 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
591 
592 	/* Path MTU Discovery Black Hole Detection related stats */
593 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
594 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
595 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
596 
597 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
598 };
599 
600 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
601 
602 #ifdef _KERNEL
603 #define	TI_UNLOCKED	1
604 #define	TI_RLOCKED	2
605 #include <sys/counter.h>
606 
607 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
608 /*
609  * In-kernel consumers can use these accessor macros directly to update
610  * stats.
611  */
612 #define	TCPSTAT_ADD(name, val)	\
613     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
614 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
615 
616 /*
617  * Kernel module consumers must use this accessor macro.
618  */
619 void	kmod_tcpstat_inc(int statnum);
620 #define	KMOD_TCPSTAT_INC(name)						\
621     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
622 
623 /*
624  * Running TCP connection count by state.
625  */
626 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
627 #define	V_tcps_states	VNET(tcps_states)
628 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
629 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
630 
631 /*
632  * TCP specific helper hook point identifiers.
633  */
634 #define	HHOOK_TCP_EST_IN		0
635 #define	HHOOK_TCP_EST_OUT		1
636 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
637 
638 struct tcp_hhook_data {
639 	struct tcpcb	*tp;
640 	struct tcphdr	*th;
641 	struct tcpopt	*to;
642 	uint32_t	len;
643 	int		tso;
644 	tcp_seq		curack;
645 };
646 #endif
647 
648 /*
649  * TCB structure exported to user-land via sysctl(3).
650  *
651  * Fields prefixed with "xt_" are unique to the export structure, and fields
652  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
653  *
654  * Legend:
655  * (s) - used by userland utilities in src
656  * (p) - used by utilities in ports
657  * (3) - is known to be used by third party software not in ports
658  * (n) - no known usage
659  *
660  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
661  * included.  Not all of our clients do.
662  */
663 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
664 struct xtcpcb {
665 	size_t		xt_len;		/* length of this structure */
666 	struct xinpcb	xt_inp;
667 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
668 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
669 	int64_t		spare64[8];
670 	int32_t		t_state;		/* (s,p) */
671 	uint32_t	t_flags;		/* (s,p) */
672 	int32_t		t_sndzerowin;		/* (s) */
673 	int32_t		t_sndrexmitpack;	/* (s) */
674 	int32_t		t_rcvoopack;		/* (s) */
675 	int32_t		t_rcvtime;		/* (s) */
676 	int32_t		tt_rexmt;		/* (s) */
677 	int32_t		tt_persist;		/* (s) */
678 	int32_t		tt_keep;		/* (s) */
679 	int32_t		tt_2msl;		/* (s) */
680 	int32_t		tt_delack;		/* (s) */
681 	int32_t		t_logstate;		/* (3) */
682 	int32_t		spare32[32];
683 } __aligned(8);
684 
685 #ifdef _KERNEL
686 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
687 #endif
688 #endif
689 
690 /*
691  * TCP function name-to-id mapping exported to user-land via sysctl(3).
692  */
693 struct tcp_function_id {
694 	uint8_t		tfi_id;
695 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
696 };
697 
698 /*
699  * Identifiers for TCP sysctl nodes
700  */
701 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
702 #define	TCPCTL_MSSDFLT		3	/* MSS default */
703 #define TCPCTL_STATS		4	/* statistics */
704 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
705 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
706 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
707 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
708 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
709 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
710 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
711 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
712 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
713 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
714 #define	TCPCTL_DROP		15	/* drop tcp connection */
715 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
716 
717 #ifdef _KERNEL
718 #ifdef SYSCTL_DECL
719 SYSCTL_DECL(_net_inet_tcp);
720 SYSCTL_DECL(_net_inet_tcp_sack);
721 MALLOC_DECLARE(M_TCPLOG);
722 #endif
723 
724 extern	int tcp_log_in_vain;
725 
726 /*
727  * Global TCP tunables shared between different stacks.
728  * Please keep the list sorted.
729  */
730 VNET_DECLARE(int, drop_synfin);
731 VNET_DECLARE(int, path_mtu_discovery);
732 VNET_DECLARE(int, tcp_abc_l_var);
733 VNET_DECLARE(int, tcp_autorcvbuf_inc);
734 VNET_DECLARE(int, tcp_autorcvbuf_max);
735 VNET_DECLARE(int, tcp_autosndbuf_inc);
736 VNET_DECLARE(int, tcp_autosndbuf_max);
737 VNET_DECLARE(int, tcp_delack_enabled);
738 VNET_DECLARE(int, tcp_do_autorcvbuf);
739 VNET_DECLARE(int, tcp_do_autosndbuf);
740 VNET_DECLARE(int, tcp_do_ecn);
741 VNET_DECLARE(int, tcp_do_rfc1323);
742 VNET_DECLARE(int, tcp_do_rfc3042);
743 VNET_DECLARE(int, tcp_do_rfc3390);
744 VNET_DECLARE(int, tcp_do_rfc3465);
745 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
746 VNET_DECLARE(int, tcp_do_sack);
747 VNET_DECLARE(int, tcp_do_tso);
748 VNET_DECLARE(int, tcp_ecn_maxretries);
749 VNET_DECLARE(int, tcp_initcwnd_segments);
750 VNET_DECLARE(int, tcp_insecure_rst);
751 VNET_DECLARE(int, tcp_insecure_syn);
752 VNET_DECLARE(int, tcp_minmss);
753 VNET_DECLARE(int, tcp_mssdflt);
754 VNET_DECLARE(int, tcp_recvspace);
755 VNET_DECLARE(int, tcp_sack_globalholes);
756 VNET_DECLARE(int, tcp_sack_globalmaxholes);
757 VNET_DECLARE(int, tcp_sack_maxholes);
758 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
759 VNET_DECLARE(int, tcp_sendspace);
760 VNET_DECLARE(struct inpcbhead, tcb);
761 VNET_DECLARE(struct inpcbinfo, tcbinfo);
762 
763 #define	V_drop_synfin			VNET(drop_synfin)
764 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
765 #define	V_tcb				VNET(tcb)
766 #define	V_tcbinfo			VNET(tcbinfo)
767 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
768 #define	V_tcp_autorcvbuf_inc		VNET(tcp_autorcvbuf_inc)
769 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
770 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
771 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
772 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
773 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
774 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
775 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
776 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
777 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
778 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
779 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
780 #define	V_tcp_do_rfc6675_pipe		VNET(tcp_do_rfc6675_pipe)
781 #define	V_tcp_do_sack			VNET(tcp_do_sack)
782 #define	V_tcp_do_tso			VNET(tcp_do_tso)
783 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
784 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
785 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
786 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
787 #define	V_tcp_minmss			VNET(tcp_minmss)
788 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
789 #define	V_tcp_recvspace			VNET(tcp_recvspace)
790 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
791 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
792 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
793 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
794 #define	V_tcp_sendspace			VNET(tcp_sendspace)
795 
796 #ifdef TCP_HHOOK
797 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
798 #define	V_tcp_hhh		VNET(tcp_hhh)
799 #endif
800 
801 int	 tcp_addoptions(struct tcpopt *, u_char *);
802 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
803 struct tcpcb *
804 	 tcp_close(struct tcpcb *);
805 void	 tcp_discardcb(struct tcpcb *);
806 void	 tcp_twstart(struct tcpcb *);
807 void	 tcp_twclose(struct tcptw *, int);
808 void	 tcp_ctlinput(int, struct sockaddr *, void *);
809 int	 tcp_ctloutput(struct socket *, struct sockopt *);
810 struct tcpcb *
811 	 tcp_drop(struct tcpcb *, int);
812 void	 tcp_drain(void);
813 void	 tcp_init(void);
814 void	 tcp_fini(void *);
815 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
816 	    const void *);
817 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
818 	    const void *);
819 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
820 void	 tcp_reass_global_init(void);
821 void	 tcp_reass_flush(struct tcpcb *);
822 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
823 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
824 		     struct tcpcb *, int, int);
825 void	tcp_pulloutofband(struct socket *,
826 		     struct tcphdr *, struct mbuf *, int);
827 void	tcp_xmit_timer(struct tcpcb *, int);
828 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
829 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
830 			    uint16_t nsegs, uint16_t type);
831 void 	cc_conn_init(struct tcpcb *tp);
832 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
833 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
834 #ifdef TCP_HHOOK
835 void	hhook_run_tcp_est_in(struct tcpcb *tp,
836 			    struct tcphdr *th, struct tcpopt *to);
837 #endif
838 
839 int	 tcp_input(struct mbuf **, int *, int);
840 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
841 	    struct tcpcb *, int);
842 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
843 			struct socket *, struct tcpcb *, int, int, uint8_t,
844 			int);
845 
846 int register_tcp_functions(struct tcp_function_block *blk, int wait);
847 int register_tcp_functions_as_names(struct tcp_function_block *blk,
848     int wait, const char *names[], int *num_names);
849 int register_tcp_functions_as_name(struct tcp_function_block *blk,
850     const char *name, int wait);
851 int deregister_tcp_functions(struct tcp_function_block *blk);
852 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
853 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
854 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
855 
856 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
857 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
858 u_int	 tcp_maxseg(const struct tcpcb *);
859 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
860 	    struct tcp_ifcap *);
861 void	 tcp_mss(struct tcpcb *, int);
862 int	 tcp_mssopt(struct in_conninfo *);
863 struct inpcb *
864 	 tcp_drop_syn_sent(struct inpcb *, int);
865 struct tcpcb *
866 	 tcp_newtcpcb(struct inpcb *);
867 int	 tcp_output(struct tcpcb *);
868 void	 tcp_state_change(struct tcpcb *, int);
869 void	 tcp_respond(struct tcpcb *, void *,
870 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
871 void	 tcp_tw_init(void);
872 #ifdef VIMAGE
873 void	 tcp_tw_destroy(void);
874 #endif
875 void	 tcp_tw_zone_change(void);
876 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
877 	    struct mbuf *, int);
878 void	 tcp_setpersist(struct tcpcb *);
879 void	 tcp_slowtimo(void);
880 struct tcptemp *
881 	 tcpip_maketemplate(struct inpcb *);
882 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
883 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
884 int	 tcp_timer_active(struct tcpcb *, uint32_t);
885 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
886 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
887 /*
888  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
889  */
890 void	 tcp_hc_init(void);
891 #ifdef VIMAGE
892 void	 tcp_hc_destroy(void);
893 #endif
894 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
895 uint32_t tcp_hc_getmtu(struct in_conninfo *);
896 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
897 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
898 
899 extern	struct pr_usrreqs tcp_usrreqs;
900 tcp_seq tcp_new_isn(struct tcpcb *);
901 
902 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
903 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
904 void	 tcp_clean_sackreport(struct tcpcb *tp);
905 void	 tcp_sack_adjust(struct tcpcb *tp);
906 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
907 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
908 void	 tcp_free_sackholes(struct tcpcb *tp);
909 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
910 int	 tcp_compute_pipe(struct tcpcb *);
911 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
912 
913 static inline void
914 tcp_fields_to_host(struct tcphdr *th)
915 {
916 
917 	th->th_seq = ntohl(th->th_seq);
918 	th->th_ack = ntohl(th->th_ack);
919 	th->th_win = ntohs(th->th_win);
920 	th->th_urp = ntohs(th->th_urp);
921 }
922 
923 static inline void
924 tcp_fields_to_net(struct tcphdr *th)
925 {
926 
927 	th->th_seq = htonl(th->th_seq);
928 	th->th_ack = htonl(th->th_ack);
929 	th->th_win = htons(th->th_win);
930 	th->th_urp = htons(th->th_urp);
931 }
932 #endif /* _KERNEL */
933 
934 #endif /* _NETINET_TCP_VAR_H_ */
935