xref: /freebsd/sys/netinet/tcp_var.h (revision 3ee5c55415a7b08c6c4c403cc6b96e30d768e1c9)
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 64 byte cacheline efficiency based
87  * on common tcp_input/tcp_output processing.
88  */
89 struct tcpcb {
90 	/* Cache line 1 */
91 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
92 	struct tcp_function_block *t_fb;/* TCP function call block */
93 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
94 	uint32_t t_maxseg:24,		/* maximum segment size */
95 		t_logstate:8;		/* State of "black box" logging */
96 	uint32_t t_state:4,		/* state of this connection */
97 		bits_spare : 24;
98 	u_int	t_flags;
99 	tcp_seq	snd_una;		/* sent but unacknowledged */
100 	tcp_seq	snd_max;		/* highest sequence number sent;
101 					 * used to recognize retransmits
102 					 */
103 	tcp_seq	snd_nxt;		/* send next */
104 	tcp_seq	snd_up;			/* send urgent pointer */
105 	uint32_t  snd_wnd;		/* send window */
106 	uint32_t  snd_cwnd;		/* congestion-controlled window */
107 	uint32_t cl1_spare; 		/* Spare to round out CL 1 */
108 	/* Cache line 2 */
109 	u_int32_t  ts_offset;		/* our timestamp offset */
110 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
111 	int	rcv_numsacks;		/* # distinct sack blks present */
112 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
113 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
114 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
115 	tcp_seq	rcv_nxt;		/* receive next */
116 	tcp_seq	rcv_adv;		/* advertised window */
117 	uint32_t  rcv_wnd;		/* receive window */
118 	u_int	t_flags2;		/* More tcpcb flags storage */
119 	int	t_srtt;			/* smoothed round-trip time */
120 	int	t_rttvar;		/* variance in round-trip time */
121 	u_int32_t  ts_recent;		/* timestamp echo data */
122 	u_char	snd_scale;		/* window scaling for send window */
123 	u_char	rcv_scale;		/* window scaling for recv window */
124 	u_char	snd_limited;		/* segments limited transmitted */
125 	u_char	request_r_scale;	/* pending window scaling */
126 	tcp_seq	last_ack_sent;
127 	u_int	t_rcvtime;		/* inactivity time */
128 	/* Cache line 3 */
129 	tcp_seq	rcv_up;			/* receive urgent pointer */
130 	int	t_segqlen;		/* segment reassembly queue length */
131 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
132 	struct mbuf      *t_in_pkt;
133 	struct mbuf	 *t_tail_pkt;
134 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
135 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
136 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
137 					 * for slow start exponential to
138 					 * linear switch
139 					 */
140 	tcp_seq	snd_wl1;		/* window update seg seq number */
141 	/* Cache line 4 */
142 	tcp_seq	snd_wl2;		/* window update seg ack number */
143 
144 	tcp_seq	irs;			/* initial receive sequence number */
145 	tcp_seq	iss;		        /* initial send sequence number */
146 	u_int   t_acktime;
147 	u_int	ts_recent_age;		/* when last updated */
148 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
149 	uint16_t cl4_spare;		/* Spare to adjust CL 4 */
150 	char	t_oobflags;		/* have some */
151 	char	t_iobc;			/* input character */
152 	int	t_rxtcur;		/* current retransmit value (ticks) */
153 
154 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
155 	u_int	t_rtttime;		/* RTT measurement start time */
156 
157 	tcp_seq	t_rtseq;		/* sequence number being timed */
158 	u_int	t_starttime;		/* time connection was established */
159 
160 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
161 	u_int	t_rttmin;		/* minimum rtt allowed */
162 
163 	u_int	t_rttbest;		/* best rtt we've seen */
164 
165 	int	t_softerror;		/* possible error not yet reported */
166 	uint32_t  max_sndwnd;		/* largest window peer has offered */
167 	/* Cache line 5 */
168 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
169 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
170 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
171 	int	t_sndzerowin;		/* zero-window updates sent */
172 	u_long	t_rttupdated;		/* number of times rtt sampled */
173 	int	snd_numholes;		/* number of holes seen by sender */
174 	u_int	t_badrxtwin;		/* window for retransmit recovery */
175 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
176 					/* SACK scoreboard (sorted) */
177 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
178 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
179 					   episode starts at this seq number */
180 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
181 	struct sackhint	sackhint;	/* SACK scoreboard hint */
182 	int	t_rttlow;		/* smallest observerved RTT */
183 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
184 	struct toedev	*tod;		/* toedev handling this connection */
185 	int	t_sndrexmitpack;	/* retransmit packets sent */
186 	int	t_rcvoopack;		/* out-of-order packets received */
187 	void	*t_toe;			/* TOE pcb pointer */
188 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
189 	struct cc_var	*ccv;		/* congestion control specific vars */
190 	struct osd	*osd;		/* storage for Khelp module data */
191 	int	t_bytes_acked;		/* # bytes acked during current RTT */
192 	u_int	t_keepinit;		/* time to establish connection */
193 	u_int	t_keepidle;		/* time before keepalive probes begin */
194 	u_int	t_keepintvl;		/* interval between keepalives */
195 	u_int	t_keepcnt;		/* number of keepalives before close */
196 	int	t_dupacks;		/* consecutive dup acks recd */
197 	int	t_lognum;		/* Number of log entries */
198 	struct tcp_log_stailq t_logs;	/* Log buffer */
199 	struct tcp_log_id_node *t_lin;
200 	struct tcp_log_id_bucket *t_lib;
201 	const char *t_output_caller;	/* Function that called tcp_output */
202 	uint32_t t_logsn;		/* Log "serial number" */
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 	int	(*tfb_tcp_output_wtime)(struct tcpcb *, const struct timeval *);
261 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
262 			    struct socket *, struct tcpcb *,
263 			    int, int, uint8_t,
264 			    int);
265 	void	(*tfb_tcp_hpts_do_segment)(struct mbuf *, struct tcphdr *,
266 			    struct socket *, struct tcpcb *,
267 			    int, int, uint8_t,
268 			    int, int, struct timeval *);
269 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
270 			    struct inpcb *inp, struct tcpcb *tp);
271 	/* Optional memory allocation/free routine */
272 	int	(*tfb_tcp_fb_init)(struct tcpcb *);
273 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
274 	/* Optional timers, must define all if you define one */
275 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
276 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
277 			    uint32_t, u_int);
278 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
279 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
280 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
281 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
282 	void	(*tfb_tcp_mtu_chg)(struct tcpcb *);
283 	volatile uint32_t tfb_refcnt;
284 	uint32_t  tfb_flags;
285 	uint8_t	tfb_id;
286 };
287 
288 struct tcp_function {
289 	TAILQ_ENTRY(tcp_function)	tf_next;
290 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
291 	struct tcp_function_block	*tf_fb;
292 };
293 
294 TAILQ_HEAD(tcp_funchead, tcp_function);
295 #endif	/* _KERNEL */
296 
297 /*
298  * Flags and utility macros for the t_flags field.
299  */
300 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
301 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
302 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
303 #define	TF_NOOPT	0x000008	/* don't use tcp options */
304 #define	TF_SENTFIN	0x000010	/* have sent FIN */
305 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
306 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
307 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
308 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
309 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
310 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
311 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
312 #define	TF_NOPUSH	0x001000	/* don't push */
313 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
314 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
315 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
316 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
317 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
318 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
319 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
320 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
321 #define	TF_FORCEDATA	0x800000	/* force out a byte */
322 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
323 #define	TF_TOE		0x2000000	/* this connection is offloaded */
324 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
325 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
326 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
327 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
328 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
329 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
330 
331 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
332 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
333 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
334 
335 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
336 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
337 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
338 
339 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
340 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
341 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
342 
343 #if defined(_KERNEL) && !defined(TCP_RFC7413)
344 #define	IS_FASTOPEN(t_flags)		(false)
345 #else
346 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
347 #endif
348 
349 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
350 
351 /*
352  * Flags for the t_oobflags field.
353  */
354 #define	TCPOOB_HAVEDATA	0x01
355 #define	TCPOOB_HADDATA	0x02
356 
357 /*
358  * Flags for the extended TCP flags field, t_flags2
359  */
360 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
361 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
362 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
363 #define	TF2_LOG_AUTO		0x00000008 /* Session is auto-logging. */
364 
365 /*
366  * Structure to hold TCP options that are only used during segment
367  * processing (in tcp_input), but not held in the tcpcb.
368  * It's basically used to reduce the number of parameters
369  * to tcp_dooptions and tcp_addoptions.
370  * The binary order of the to_flags is relevant for packing of the
371  * options in tcp_addoptions.
372  */
373 struct tcpopt {
374 	u_int32_t	to_flags;	/* which options are present */
375 #define	TOF_MSS		0x0001		/* maximum segment size */
376 #define	TOF_SCALE	0x0002		/* window scaling */
377 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
378 #define	TOF_TS		0x0010		/* timestamp */
379 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
380 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
381 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
382 #define	TOF_MAXOPT	0x0200
383 	u_int32_t	to_tsval;	/* new timestamp */
384 	u_int32_t	to_tsecr;	/* reflected timestamp */
385 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
386 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
387 	u_int8_t	*to_tfo_cookie; /* pointer to the TFO cookie */
388 	u_int16_t	to_mss;		/* maximum segment size */
389 	u_int8_t	to_wscale;	/* window scaling */
390 	u_int8_t	to_nsacks;	/* number of SACK blocks */
391 	u_int8_t	to_tfo_len;	/* TFO cookie length */
392 	u_int32_t	to_spare;	/* UTO */
393 };
394 
395 /*
396  * Flags for tcp_dooptions.
397  */
398 #define	TO_SYN		0x01		/* parse SYN-only options */
399 
400 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
401 	uint32_t	rmx_mtu;	/* MTU for this path */
402 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
403 	uint32_t	rmx_rtt;	/* estimated round trip time */
404 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
405 	uint32_t	rmx_cwnd;	/* congestion window */
406 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
407 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
408 };
409 
410 /*
411  * Used by tcp_maxmtu() to communicate interface specific features
412  * and limits at the time of connection setup.
413  */
414 struct tcp_ifcap {
415 	int	ifcap;
416 	u_int	tsomax;
417 	u_int	tsomaxsegcount;
418 	u_int	tsomaxsegsize;
419 };
420 
421 #ifndef _NETINET_IN_PCB_H_
422 struct in_conninfo;
423 #endif /* _NETINET_IN_PCB_H_ */
424 
425 struct tcptw {
426 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
427 	tcp_seq		snd_nxt;
428 	tcp_seq		rcv_nxt;
429 	tcp_seq		iss;
430 	tcp_seq		irs;
431 	u_short		last_win;	/* cached window value */
432 	short		tw_so_options;	/* copy of so_options */
433 	struct ucred	*tw_cred;	/* user credentials */
434 	u_int32_t	t_recent;
435 	u_int32_t	ts_offset;	/* our timestamp offset */
436 	u_int		t_starttime;
437 	int		tw_time;
438 	TAILQ_ENTRY(tcptw) tw_2msl;
439 	void		*tw_pspare;	/* TCP_SIGNATURE */
440 	u_int		*tw_spare;	/* TCP_SIGNATURE */
441 };
442 
443 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
444 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
445 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
446 
447 /*
448  * The smoothed round-trip time and estimated variance
449  * are stored as fixed point numbers scaled by the values below.
450  * For convenience, these scales are also used in smoothing the average
451  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
452  * With these scales, srtt has 3 bits to the right of the binary point,
453  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
454  * binary point, and is smoothed with an ALPHA of 0.75.
455  */
456 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
457 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
458 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
459 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
460 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
461 
462 /*
463  * The initial retransmission should happen at rtt + 4 * rttvar.
464  * Because of the way we do the smoothing, srtt and rttvar
465  * will each average +1/2 tick of bias.  When we compute
466  * the retransmit timer, we want 1/2 tick of rounding and
467  * 1 extra tick because of +-1/2 tick uncertainty in the
468  * firing of the timer.  The bias will give us exactly the
469  * 1.5 tick we need.  But, because the bias is
470  * statistical, we have to test that we don't drop below
471  * the minimum feasible timer (which is 2 ticks).
472  * This version of the macro adapted from a paper by Lawrence
473  * Brakmo and Larry Peterson which outlines a problem caused
474  * by insufficient precision in the original implementation,
475  * which results in inappropriately large RTO values for very
476  * fast networks.
477  */
478 #define	TCP_REXMTVAL(tp) \
479 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
480 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
481 
482 /*
483  * TCP statistics.
484  * Many of these should be kept per connection,
485  * but that's inconvenient at the moment.
486  */
487 struct	tcpstat {
488 	uint64_t tcps_connattempt;	/* connections initiated */
489 	uint64_t tcps_accepts;		/* connections accepted */
490 	uint64_t tcps_connects;		/* connections established */
491 	uint64_t tcps_drops;		/* connections dropped */
492 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
493 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
494 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
495 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
496 	uint64_t tcps_rttupdated;	/* times we succeeded */
497 	uint64_t tcps_delack;		/* delayed acks sent */
498 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
499 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
500 	uint64_t tcps_persisttimeo;	/* persist timeouts */
501 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
502 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
503 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
504 
505 	uint64_t tcps_sndtotal;		/* total packets sent */
506 	uint64_t tcps_sndpack;		/* data packets sent */
507 	uint64_t tcps_sndbyte;		/* data bytes sent */
508 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
509 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
510 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
511 	uint64_t tcps_sndacks;		/* ack-only packets sent */
512 	uint64_t tcps_sndprobe;		/* window probes sent */
513 	uint64_t tcps_sndurg;		/* packets sent with URG only */
514 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
515 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
516 
517 	uint64_t tcps_rcvtotal;		/* total packets received */
518 	uint64_t tcps_rcvpack;		/* packets received in sequence */
519 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
520 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
521 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
522 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
523 	uint64_t tcps_rcvshort;		/* packets received too short */
524 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
525 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
526 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
527 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
528 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
529 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
530 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
531 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
532 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
533 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
534 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
535 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
536 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
537 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
538 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
539 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
540 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
541 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
542 	uint64_t tcps_pcbcachemiss;
543 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
544 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
545 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
546 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
547 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
548 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
549 	uint64_t tcps_persistdrop;	/* timeout in persist state */
550 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
551 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
552 	uint64_t tcps_listendrop;	/* listen queue overflows */
553 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
554 
555 	uint64_t tcps_sc_added;		/* entry added to syncache */
556 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
557 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
558 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
559 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
560 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
561 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
562 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
563 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
564 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
565 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
566 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
567 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
568 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
569 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
570 
571 	uint64_t tcps_hc_added;		/* entry added to hostcache */
572 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
573 
574 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
575 
576 	/* SACK related stats */
577 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
578 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
579 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
580 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
581 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
582 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
583 
584 	/* ECN related stats */
585 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
586 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
587 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
588 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
589 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
590 
591 	/* TCP_SIGNATURE related stats */
592 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
593 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
594 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
595 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
596 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
597 
598 	/* Path MTU Discovery Black Hole Detection related stats */
599 	uint64_t tcps_pmtud_blackhole_activated;	 /* Black Hole Count */
600 	uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
601 	uint64_t tcps_pmtud_blackhole_failed;		 /* Black Hole Failure Count */
602 
603 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
604 };
605 
606 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
607 
608 #ifdef _KERNEL
609 #define	TI_UNLOCKED	1
610 #define	TI_RLOCKED	2
611 #include <sys/counter.h>
612 
613 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
614 /*
615  * In-kernel consumers can use these accessor macros directly to update
616  * stats.
617  */
618 #define	TCPSTAT_ADD(name, val)	\
619     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
620 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
621 
622 /*
623  * Kernel module consumers must use this accessor macro.
624  */
625 void	kmod_tcpstat_inc(int statnum);
626 #define	KMOD_TCPSTAT_INC(name)						\
627     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
628 
629 /*
630  * Running TCP connection count by state.
631  */
632 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
633 #define	V_tcps_states	VNET(tcps_states)
634 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
635 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
636 
637 /*
638  * TCP specific helper hook point identifiers.
639  */
640 #define	HHOOK_TCP_EST_IN		0
641 #define	HHOOK_TCP_EST_OUT		1
642 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
643 
644 struct tcp_hhook_data {
645 	struct tcpcb	*tp;
646 	struct tcphdr	*th;
647 	struct tcpopt	*to;
648 	uint32_t	len;
649 	int		tso;
650 	tcp_seq		curack;
651 };
652 #endif
653 
654 /*
655  * TCB structure exported to user-land via sysctl(3).
656  *
657  * Fields prefixed with "xt_" are unique to the export structure, and fields
658  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
659  *
660  * Legend:
661  * (s) - used by userland utilities in src
662  * (p) - used by utilities in ports
663  * (3) - is known to be used by third party software not in ports
664  * (n) - no known usage
665  *
666  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
667  * included.  Not all of our clients do.
668  */
669 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
670 struct xtcpcb {
671 	size_t		xt_len;		/* length of this structure */
672 	struct xinpcb	xt_inp;
673 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (s) */
674 	char		xt_logid[TCP_LOG_ID_LEN];	/* (s) */
675 	int64_t		spare64[8];
676 	int32_t		t_state;		/* (s,p) */
677 	uint32_t	t_flags;		/* (s,p) */
678 	int32_t		t_sndzerowin;		/* (s) */
679 	int32_t		t_sndrexmitpack;	/* (s) */
680 	int32_t		t_rcvoopack;		/* (s) */
681 	int32_t		t_rcvtime;		/* (s) */
682 	int32_t		tt_rexmt;		/* (s) */
683 	int32_t		tt_persist;		/* (s) */
684 	int32_t		tt_keep;		/* (s) */
685 	int32_t		tt_2msl;		/* (s) */
686 	int32_t		tt_delack;		/* (s) */
687 	int32_t		t_logstate;		/* (3) */
688 	int32_t		spare32[32];
689 } __aligned(8);
690 
691 #ifdef _KERNEL
692 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
693 #endif
694 #endif
695 
696 /*
697  * TCP function information (name-to-id mapping, aliases, and refcnt)
698  * exported to user-land via sysctl(3).
699  */
700 struct tcp_function_info {
701 	uint32_t	tfi_refcnt;
702 	uint8_t		tfi_id;
703 	char		tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
704 	char		tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
705 };
706 
707 /*
708  * Identifiers for TCP sysctl nodes
709  */
710 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
711 #define	TCPCTL_MSSDFLT		3	/* MSS default */
712 #define TCPCTL_STATS		4	/* statistics */
713 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
714 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
715 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
716 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
717 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
718 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
719 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
720 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
721 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
722 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
723 #define	TCPCTL_DROP		15	/* drop tcp connection */
724 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
725 
726 #ifdef _KERNEL
727 #ifdef SYSCTL_DECL
728 SYSCTL_DECL(_net_inet_tcp);
729 SYSCTL_DECL(_net_inet_tcp_sack);
730 MALLOC_DECLARE(M_TCPLOG);
731 #endif
732 
733 extern	int tcp_log_in_vain;
734 
735 /*
736  * Global TCP tunables shared between different stacks.
737  * Please keep the list sorted.
738  */
739 VNET_DECLARE(int, drop_synfin);
740 VNET_DECLARE(int, path_mtu_discovery);
741 VNET_DECLARE(int, tcp_abc_l_var);
742 VNET_DECLARE(int, tcp_autorcvbuf_inc);
743 VNET_DECLARE(int, tcp_autorcvbuf_max);
744 VNET_DECLARE(int, tcp_autosndbuf_inc);
745 VNET_DECLARE(int, tcp_autosndbuf_max);
746 VNET_DECLARE(int, tcp_delack_enabled);
747 VNET_DECLARE(int, tcp_do_autorcvbuf);
748 VNET_DECLARE(int, tcp_do_autosndbuf);
749 VNET_DECLARE(int, tcp_do_ecn);
750 VNET_DECLARE(int, tcp_do_rfc1323);
751 VNET_DECLARE(int, tcp_do_rfc3042);
752 VNET_DECLARE(int, tcp_do_rfc3390);
753 VNET_DECLARE(int, tcp_do_rfc3465);
754 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
755 VNET_DECLARE(int, tcp_do_sack);
756 VNET_DECLARE(int, tcp_do_tso);
757 VNET_DECLARE(int, tcp_ecn_maxretries);
758 VNET_DECLARE(int, tcp_initcwnd_segments);
759 VNET_DECLARE(int, tcp_insecure_rst);
760 VNET_DECLARE(int, tcp_insecure_syn);
761 VNET_DECLARE(int, tcp_minmss);
762 VNET_DECLARE(int, tcp_mssdflt);
763 VNET_DECLARE(int, tcp_recvspace);
764 VNET_DECLARE(int, tcp_sack_globalholes);
765 VNET_DECLARE(int, tcp_sack_globalmaxholes);
766 VNET_DECLARE(int, tcp_sack_maxholes);
767 VNET_DECLARE(int, tcp_sc_rst_sock_fail);
768 VNET_DECLARE(int, tcp_sendspace);
769 VNET_DECLARE(struct inpcbhead, tcb);
770 VNET_DECLARE(struct inpcbinfo, tcbinfo);
771 
772 #define	V_drop_synfin			VNET(drop_synfin)
773 #define	V_path_mtu_discovery		VNET(path_mtu_discovery)
774 #define	V_tcb				VNET(tcb)
775 #define	V_tcbinfo			VNET(tcbinfo)
776 #define	V_tcp_abc_l_var			VNET(tcp_abc_l_var)
777 #define	V_tcp_autorcvbuf_inc		VNET(tcp_autorcvbuf_inc)
778 #define	V_tcp_autorcvbuf_max		VNET(tcp_autorcvbuf_max)
779 #define	V_tcp_autosndbuf_inc		VNET(tcp_autosndbuf_inc)
780 #define	V_tcp_autosndbuf_max		VNET(tcp_autosndbuf_max)
781 #define	V_tcp_delack_enabled		VNET(tcp_delack_enabled)
782 #define	V_tcp_do_autorcvbuf		VNET(tcp_do_autorcvbuf)
783 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
784 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
785 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
786 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
787 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
788 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
789 #define	V_tcp_do_rfc6675_pipe		VNET(tcp_do_rfc6675_pipe)
790 #define	V_tcp_do_sack			VNET(tcp_do_sack)
791 #define	V_tcp_do_tso			VNET(tcp_do_tso)
792 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)
793 #define	V_tcp_initcwnd_segments		VNET(tcp_initcwnd_segments)
794 #define	V_tcp_insecure_rst		VNET(tcp_insecure_rst)
795 #define	V_tcp_insecure_syn		VNET(tcp_insecure_syn)
796 #define	V_tcp_minmss			VNET(tcp_minmss)
797 #define	V_tcp_mssdflt			VNET(tcp_mssdflt)
798 #define	V_tcp_recvspace			VNET(tcp_recvspace)
799 #define	V_tcp_sack_globalholes		VNET(tcp_sack_globalholes)
800 #define	V_tcp_sack_globalmaxholes	VNET(tcp_sack_globalmaxholes)
801 #define	V_tcp_sack_maxholes		VNET(tcp_sack_maxholes)
802 #define	V_tcp_sc_rst_sock_fail		VNET(tcp_sc_rst_sock_fail)
803 #define	V_tcp_sendspace			VNET(tcp_sendspace)
804 
805 #ifdef TCP_HHOOK
806 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
807 #define	V_tcp_hhh		VNET(tcp_hhh)
808 #endif
809 
810 int	 tcp_addoptions(struct tcpopt *, u_char *);
811 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
812 struct tcpcb *
813 	 tcp_close(struct tcpcb *);
814 void	 tcp_discardcb(struct tcpcb *);
815 void	 tcp_twstart(struct tcpcb *);
816 void	 tcp_twclose(struct tcptw *, int);
817 void	 tcp_ctlinput(int, struct sockaddr *, void *);
818 int	 tcp_ctloutput(struct socket *, struct sockopt *);
819 struct tcpcb *
820 	 tcp_drop(struct tcpcb *, int);
821 void	 tcp_drain(void);
822 void	 tcp_init(void);
823 void	 tcp_fini(void *);
824 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
825 	    const void *);
826 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
827 	    const void *);
828 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
829 void	 tcp_reass_global_init(void);
830 void	 tcp_reass_flush(struct tcpcb *);
831 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
832 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
833 		     struct tcpcb *, int, int);
834 void	tcp_pulloutofband(struct socket *,
835 		     struct tcphdr *, struct mbuf *, int);
836 void	tcp_xmit_timer(struct tcpcb *, int);
837 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
838 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
839 			    uint16_t nsegs, uint16_t type);
840 void 	cc_conn_init(struct tcpcb *tp);
841 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
842 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
843 #ifdef TCP_HHOOK
844 void	hhook_run_tcp_est_in(struct tcpcb *tp,
845 			    struct tcphdr *th, struct tcpopt *to);
846 #endif
847 
848 int	 tcp_input(struct mbuf **, int *, int);
849 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
850 	    struct tcpcb *, int);
851 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
852 			struct socket *, struct tcpcb *, int, int, uint8_t,
853 			int);
854 
855 int register_tcp_functions(struct tcp_function_block *blk, int wait);
856 int register_tcp_functions_as_names(struct tcp_function_block *blk,
857     int wait, const char *names[], int *num_names);
858 int register_tcp_functions_as_name(struct tcp_function_block *blk,
859     const char *name, int wait);
860 int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
861     bool force);
862 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
863 void tcp_switch_back_to_default(struct tcpcb *tp);
864 struct tcp_function_block *
865 find_and_ref_tcp_fb(struct tcp_function_block *fs);
866 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
867 
868 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
869 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
870 u_int	 tcp_maxseg(const struct tcpcb *);
871 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
872 	    struct tcp_ifcap *);
873 void	 tcp_mss(struct tcpcb *, int);
874 int	 tcp_mssopt(struct in_conninfo *);
875 struct inpcb *
876 	 tcp_drop_syn_sent(struct inpcb *, int);
877 struct tcpcb *
878 	 tcp_newtcpcb(struct inpcb *);
879 int	 tcp_output(struct tcpcb *);
880 void	 tcp_state_change(struct tcpcb *, int);
881 void	 tcp_respond(struct tcpcb *, void *,
882 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
883 void	 tcp_tw_init(void);
884 #ifdef VIMAGE
885 void	 tcp_tw_destroy(void);
886 #endif
887 void	 tcp_tw_zone_change(void);
888 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
889 	    struct mbuf *, int);
890 void	 tcp_setpersist(struct tcpcb *);
891 void	 tcp_slowtimo(void);
892 struct tcptemp *
893 	 tcpip_maketemplate(struct inpcb *);
894 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
895 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
896 int	 tcp_timer_active(struct tcpcb *, uint32_t);
897 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
898 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
899 /*
900  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
901  */
902 void	 tcp_hc_init(void);
903 #ifdef VIMAGE
904 void	 tcp_hc_destroy(void);
905 #endif
906 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
907 uint32_t tcp_hc_getmtu(struct in_conninfo *);
908 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
909 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
910 
911 extern	struct pr_usrreqs tcp_usrreqs;
912 tcp_seq tcp_new_isn(struct tcpcb *);
913 
914 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
915 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
916 void	 tcp_clean_sackreport(struct tcpcb *tp);
917 void	 tcp_sack_adjust(struct tcpcb *tp);
918 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
919 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
920 void	 tcp_free_sackholes(struct tcpcb *tp);
921 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
922 int	 tcp_compute_pipe(struct tcpcb *);
923 void	 tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
924 
925 static inline void
926 tcp_fields_to_host(struct tcphdr *th)
927 {
928 
929 	th->th_seq = ntohl(th->th_seq);
930 	th->th_ack = ntohl(th->th_ack);
931 	th->th_win = ntohs(th->th_win);
932 	th->th_urp = ntohs(th->th_urp);
933 }
934 
935 static inline void
936 tcp_fields_to_net(struct tcphdr *th)
937 {
938 
939 	th->th_seq = htonl(th->th_seq);
940 	th->th_ack = htonl(th->th_ack);
941 	th->th_win = htons(th->th_win);
942 	th->th_urp = htons(th->th_urp);
943 }
944 #endif /* _KERNEL */
945 
946 #endif /* _NETINET_TCP_VAR_H_ */
947