xref: /freebsd/sys/netinet/tcp_var.h (revision d93a896ef95946b0bf1219866fcb324b78543444)
1 /*-
2  * Copyright (c) 1982, 1986, 1993, 1994, 1995
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
30  * $FreeBSD$
31  */
32 
33 #ifndef _NETINET_TCP_VAR_H_
34 #define _NETINET_TCP_VAR_H_
35 
36 #include <netinet/tcp.h>
37 #include <netinet/tcp_fsm.h>
38 
39 #ifdef _KERNEL
40 #include <net/vnet.h>
41 #include <sys/mbuf.h>
42 #endif
43 
44 #if defined(_KERNEL) || defined(_WANT_TCPCB)
45 /* TCP segment queue entry */
46 struct tseg_qent {
47 	LIST_ENTRY(tseg_qent) tqe_q;
48 	int	tqe_len;		/* TCP segment data length */
49 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
50 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
51 };
52 LIST_HEAD(tsegqe_head, tseg_qent);
53 
54 struct sackblk {
55 	tcp_seq start;		/* start seq no. of sack block */
56 	tcp_seq end;		/* end seq no. */
57 };
58 
59 struct sackhole {
60 	tcp_seq start;		/* start seq no. of hole */
61 	tcp_seq end;		/* end seq no. */
62 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
63 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
64 };
65 
66 struct sackhint {
67 	struct sackhole	*nexthole;
68 	int		sack_bytes_rexmit;
69 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
70 
71 	int		ispare;		/* explicit pad for 64bit alignment */
72 	int             sacked_bytes;	/*
73 					 * Total sacked bytes reported by the
74 					 * receiver via sack option
75 					 */
76 	uint32_t	_pad1[1];	/* TBD */
77 	uint64_t	_pad[1];	/* TBD */
78 };
79 
80 /*
81  * Tcp control block, one per tcp; fields:
82  * Organized for 16 byte cacheline efficiency.
83  */
84 struct tcpcb {
85 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
86 	int	t_segqlen;		/* segment reassembly queue length */
87 	int	t_dupacks;		/* consecutive dup acks recd */
88 
89 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
90 
91 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
92 	int	t_state;		/* state of this connection */
93 	u_int	t_flags;
94 
95 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
96 
97 	tcp_seq	snd_una;		/* sent but unacknowledged */
98 	tcp_seq	snd_max;		/* highest sequence number sent;
99 					 * used to recognize retransmits
100 					 */
101 	tcp_seq	snd_nxt;		/* send next */
102 	tcp_seq	snd_up;			/* send urgent pointer */
103 
104 	tcp_seq	snd_wl1;		/* window update seg seq number */
105 	tcp_seq	snd_wl2;		/* window update seg ack number */
106 	tcp_seq	iss;			/* initial send sequence number */
107 	tcp_seq	irs;			/* initial receive sequence number */
108 
109 	tcp_seq	rcv_nxt;		/* receive next */
110 	tcp_seq	rcv_adv;		/* advertised window */
111 	uint32_t  rcv_wnd;		/* receive window */
112 	tcp_seq	rcv_up;			/* receive urgent pointer */
113 
114 	uint32_t  snd_wnd;		/* send window */
115 	uint32_t  snd_cwnd;		/* congestion-controlled window */
116 	uint32_t  snd_ssthresh;		/* snd_cwnd size threshold for
117 					 * for slow start exponential to
118 					 * linear switch
119 					 */
120 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
121 
122 	u_int	t_rcvtime;		/* inactivity time */
123 	u_int	t_starttime;		/* time connection was established */
124 	u_int	t_rtttime;		/* RTT measurement start time */
125 	tcp_seq	t_rtseq;		/* sequence number being timed */
126 
127 	int	t_rxtcur;		/* current retransmit value (ticks) */
128 	u_int	t_maxseg;		/* maximum segment size */
129 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
130 	int	t_srtt;			/* smoothed round-trip time */
131 	int	t_rttvar;		/* variance in round-trip time */
132 
133 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
134 	u_int	t_rttmin;		/* minimum rtt allowed */
135 	u_int	t_rttbest;		/* best rtt we've seen */
136 	u_long	t_rttupdated;		/* number of times rtt sampled */
137 	uint32_t  max_sndwnd;		/* largest window peer has offered */
138 
139 	int	t_softerror;		/* possible error not yet reported */
140 /* out-of-band data */
141 	char	t_oobflags;		/* have some */
142 	char	t_iobc;			/* input character */
143 /* RFC 1323 variables */
144 	u_char	snd_scale;		/* window scaling for send window */
145 	u_char	rcv_scale;		/* window scaling for recv window */
146 	u_char	request_r_scale;	/* pending window scaling */
147 	u_int32_t  ts_recent;		/* timestamp echo data */
148 	u_int	ts_recent_age;		/* when last updated */
149 	u_int32_t  ts_offset;		/* our timestamp offset */
150 
151 	tcp_seq	last_ack_sent;
152 /* experimental */
153 	uint32_t  snd_cwnd_prev;	/* cwnd prior to retransmit */
154 	uint32_t  snd_ssthresh_prev;	/* ssthresh prior to retransmit */
155 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
156 	int	t_sndzerowin;		/* zero-window updates sent */
157 	u_int	t_badrxtwin;		/* window for retransmit recovery */
158 	u_char	snd_limited;		/* segments limited transmitted */
159 /* SACK related state */
160 	int	snd_numholes;		/* number of holes seen by sender */
161 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
162 					/* SACK scoreboard (sorted) */
163 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
164 	int	rcv_numsacks;		/* # distinct sack blks present */
165 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
166 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
167 					   episode starts at this seq number */
168 	struct sackhint	sackhint;	/* SACK scoreboard hint */
169 	int	t_rttlow;		/* smallest observerved RTT */
170 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
171 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
172 	struct toedev	*tod;		/* toedev handling this connection */
173 	int	t_sndrexmitpack;	/* retransmit packets sent */
174 	int	t_rcvoopack;		/* out-of-order packets received */
175 	void	*t_toe;			/* TOE pcb pointer */
176 	int	t_bytes_acked;		/* # bytes acked during current RTT */
177 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
178 	struct cc_var	*ccv;		/* congestion control specific vars */
179 	struct osd	*osd;		/* storage for Khelp module data */
180 
181 	u_int	t_keepinit;		/* time to establish connection */
182 	u_int	t_keepidle;		/* time before keepalive probes begin */
183 	u_int	t_keepintvl;		/* interval between keepalives */
184 	u_int	t_keepcnt;		/* number of keepalives before close */
185 
186 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
187 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
188 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
189 	u_int	t_flags2;		/* More tcpcb flags storage */
190 	struct tcp_function_block *t_fb;/* TCP function call block */
191 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
192 #ifdef TCP_RFC7413
193 	uint64_t t_tfo_cookie;		/* TCP Fast Open cookie */
194 	unsigned int *t_tfo_pending;	/* TCP Fast Open pending counter */
195 #endif
196 #ifdef TCPPCAP
197 	struct mbufq t_inpkts;		/* List of saved input packets. */
198 	struct mbufq t_outpkts;		/* List of saved output packets. */
199 #endif
200 };
201 #endif	/* _KERNEL || _WANT_TCPCB */
202 
203 #ifdef _KERNEL
204 /*
205  * Kernel variables for tcp.
206  */
207 VNET_DECLARE(int, tcp_do_rfc1323);
208 #define	V_tcp_do_rfc1323	VNET(tcp_do_rfc1323)
209 
210 struct tcptemp {
211 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
212 	struct	tcphdr tt_t;
213 };
214 
215 /*
216  * TODO: We yet need to brave plowing in
217  * to tcp_input() and the pru_usrreq() block.
218  * Right now these go to the old standards which
219  * are somewhat ok, but in the long term may
220  * need to be changed. If we do tackle tcp_input()
221  * then we need to get rid of the tcp_do_segment()
222  * function below.
223  */
224 /* Flags for tcp functions */
225 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
226 
227 /*
228  * If defining the optional tcp_timers, in the
229  * tfb_tcp_timer_stop call you must use the
230  * callout_async_drain() function with the
231  * tcp_timer_discard callback. You should check
232  * the return of callout_async_drain() and if 0
233  * increment tt_draincnt. Since the timer sub-system
234  * does not know your callbacks you must provide a
235  * stop_all function that loops through and calls
236  * tcp_timer_stop() with each of your defined timers.
237  * Adding a tfb_tcp_handoff_ok function allows the socket
238  * option to change stacks to query you even if the
239  * connection is in a later stage. You return 0 to
240  * say you can take over and run your stack, you return
241  * non-zero (an error number) to say no you can't.
242  * If the function is undefined you can only change
243  * in the early states (before connect or listen).
244  * tfb_tcp_fb_fini is changed to add a flag to tell
245  * the old stack if the tcb is being destroyed or
246  * not. A one in the flag means the TCB is being
247  * destroyed, a zero indicates its transitioning to
248  * another stack (via socket option).
249  */
250 struct tcp_function_block {
251 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
252 	int	(*tfb_tcp_output)(struct tcpcb *);
253 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
254 			    struct socket *, struct tcpcb *,
255 			    int, int, uint8_t,
256 			    int);
257 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
258 			    struct inpcb *inp, struct tcpcb *tp);
259 	/* Optional memory allocation/free routine */
260 	void	(*tfb_tcp_fb_init)(struct tcpcb *);
261 	void	(*tfb_tcp_fb_fini)(struct tcpcb *, int);
262 	/* Optional timers, must define all if you define one */
263 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
264 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
265 			    uint32_t, u_int);
266 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
267 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
268 	void	(*tfb_tcp_rexmit_tmr)(struct tcpcb *);
269 	int	(*tfb_tcp_handoff_ok)(struct tcpcb *);
270 	volatile uint32_t tfb_refcnt;
271 	uint32_t  tfb_flags;
272 };
273 
274 struct tcp_function {
275 	TAILQ_ENTRY(tcp_function)	tf_next;
276 	char				tf_name[TCP_FUNCTION_NAME_LEN_MAX];
277 	struct tcp_function_block	*tf_fb;
278 };
279 
280 TAILQ_HEAD(tcp_funchead, tcp_function);
281 #endif	/* _KERNEL */
282 
283 /*
284  * Flags and utility macros for the t_flags field.
285  */
286 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
287 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
288 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
289 #define	TF_NOOPT	0x000008	/* don't use tcp options */
290 #define	TF_SENTFIN	0x000010	/* have sent FIN */
291 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
292 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
293 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
294 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
295 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
296 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
297 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
298 #define	TF_NOPUSH	0x001000	/* don't push */
299 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
300 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
301 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
302 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
303 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
304 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
305 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
306 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
307 #define	TF_FORCEDATA	0x800000	/* force out a byte */
308 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
309 #define	TF_TOE		0x2000000	/* this connection is offloaded */
310 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
311 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
312 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
313 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
314 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
315 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
316 
317 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
318 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
319 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
320 
321 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
322 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
323 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
324 
325 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
326 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
327 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
328 
329 #if defined(_KERNEL) && !defined(TCP_RFC7413)
330 #define	IS_FASTOPEN(t_flags)		(false)
331 #else
332 #define	IS_FASTOPEN(t_flags)		(t_flags & TF_FASTOPEN)
333 #endif
334 
335 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
336 
337 /*
338  * Flags for the t_oobflags field.
339  */
340 #define	TCPOOB_HAVEDATA	0x01
341 #define	TCPOOB_HADDATA	0x02
342 
343 /*
344  * Flags for PLPMTU handling, t_flags2
345  */
346 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
347 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
348 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
349 
350 /*
351  * Structure to hold TCP options that are only used during segment
352  * processing (in tcp_input), but not held in the tcpcb.
353  * It's basically used to reduce the number of parameters
354  * to tcp_dooptions and tcp_addoptions.
355  * The binary order of the to_flags is relevant for packing of the
356  * options in tcp_addoptions.
357  */
358 struct tcpopt {
359 	u_int32_t	to_flags;	/* which options are present */
360 #define	TOF_MSS		0x0001		/* maximum segment size */
361 #define	TOF_SCALE	0x0002		/* window scaling */
362 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
363 #define	TOF_TS		0x0010		/* timestamp */
364 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
365 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
366 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
367 #define	TOF_MAXOPT	0x0200
368 	u_int32_t	to_tsval;	/* new timestamp */
369 	u_int32_t	to_tsecr;	/* reflected timestamp */
370 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
371 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
372 	u_char		*to_tfo_cookie; /* pointer to the TFO cookie */
373 	u_int16_t	to_mss;		/* maximum segment size */
374 	u_int8_t	to_wscale;	/* window scaling */
375 	u_int8_t	to_nsacks;	/* number of SACK blocks */
376 	u_int8_t	to_tfo_len;	/* TFO cookie length */
377 	u_int32_t	to_spare;	/* UTO */
378 };
379 
380 /*
381  * Flags for tcp_dooptions.
382  */
383 #define	TO_SYN		0x01		/* parse SYN-only options */
384 
385 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
386 	uint32_t	rmx_mtu;	/* MTU for this path */
387 	uint32_t	rmx_ssthresh;	/* outbound gateway buffer limit */
388 	uint32_t	rmx_rtt;	/* estimated round trip time */
389 	uint32_t	rmx_rttvar;	/* estimated rtt variance */
390 	uint32_t	rmx_cwnd;	/* congestion window */
391 	uint32_t	rmx_sendpipe;   /* outbound delay-bandwidth product */
392 	uint32_t	rmx_recvpipe;   /* inbound delay-bandwidth product */
393 };
394 
395 /*
396  * Used by tcp_maxmtu() to communicate interface specific features
397  * and limits at the time of connection setup.
398  */
399 struct tcp_ifcap {
400 	int	ifcap;
401 	u_int	tsomax;
402 	u_int	tsomaxsegcount;
403 	u_int	tsomaxsegsize;
404 };
405 
406 #ifndef _NETINET_IN_PCB_H_
407 struct in_conninfo;
408 #endif /* _NETINET_IN_PCB_H_ */
409 
410 struct tcptw {
411 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
412 	tcp_seq		snd_nxt;
413 	tcp_seq		rcv_nxt;
414 	tcp_seq		iss;
415 	tcp_seq		irs;
416 	u_short		last_win;	/* cached window value */
417 	short		tw_so_options;	/* copy of so_options */
418 	struct ucred	*tw_cred;	/* user credentials */
419 	u_int32_t	t_recent;
420 	u_int32_t	ts_offset;	/* our timestamp offset */
421 	u_int		t_starttime;
422 	int		tw_time;
423 	TAILQ_ENTRY(tcptw) tw_2msl;
424 	void		*tw_pspare;	/* TCP_SIGNATURE */
425 	u_int		*tw_spare;	/* TCP_SIGNATURE */
426 };
427 
428 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
429 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
430 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
431 
432 /*
433  * The smoothed round-trip time and estimated variance
434  * are stored as fixed point numbers scaled by the values below.
435  * For convenience, these scales are also used in smoothing the average
436  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
437  * With these scales, srtt has 3 bits to the right of the binary point,
438  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
439  * binary point, and is smoothed with an ALPHA of 0.75.
440  */
441 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
442 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
443 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
444 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
445 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
446 
447 /*
448  * The initial retransmission should happen at rtt + 4 * rttvar.
449  * Because of the way we do the smoothing, srtt and rttvar
450  * will each average +1/2 tick of bias.  When we compute
451  * the retransmit timer, we want 1/2 tick of rounding and
452  * 1 extra tick because of +-1/2 tick uncertainty in the
453  * firing of the timer.  The bias will give us exactly the
454  * 1.5 tick we need.  But, because the bias is
455  * statistical, we have to test that we don't drop below
456  * the minimum feasible timer (which is 2 ticks).
457  * This version of the macro adapted from a paper by Lawrence
458  * Brakmo and Larry Peterson which outlines a problem caused
459  * by insufficient precision in the original implementation,
460  * which results in inappropriately large RTO values for very
461  * fast networks.
462  */
463 #define	TCP_REXMTVAL(tp) \
464 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
465 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
466 
467 /*
468  * TCP statistics.
469  * Many of these should be kept per connection,
470  * but that's inconvenient at the moment.
471  */
472 struct	tcpstat {
473 	uint64_t tcps_connattempt;	/* connections initiated */
474 	uint64_t tcps_accepts;		/* connections accepted */
475 	uint64_t tcps_connects;		/* connections established */
476 	uint64_t tcps_drops;		/* connections dropped */
477 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
478 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
479 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
480 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
481 	uint64_t tcps_rttupdated;	/* times we succeeded */
482 	uint64_t tcps_delack;		/* delayed acks sent */
483 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
484 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
485 	uint64_t tcps_persisttimeo;	/* persist timeouts */
486 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
487 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
488 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
489 
490 	uint64_t tcps_sndtotal;		/* total packets sent */
491 	uint64_t tcps_sndpack;		/* data packets sent */
492 	uint64_t tcps_sndbyte;		/* data bytes sent */
493 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
494 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
495 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
496 	uint64_t tcps_sndacks;		/* ack-only packets sent */
497 	uint64_t tcps_sndprobe;		/* window probes sent */
498 	uint64_t tcps_sndurg;		/* packets sent with URG only */
499 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
500 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
501 
502 	uint64_t tcps_rcvtotal;		/* total packets received */
503 	uint64_t tcps_rcvpack;		/* packets received in sequence */
504 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
505 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
506 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
507 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
508 	uint64_t tcps_rcvshort;		/* packets received too short */
509 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
510 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
511 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
512 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
513 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
514 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
515 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
516 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
517 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
518 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
519 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
520 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
521 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
522 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
523 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
524 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
525 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
526 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
527 	uint64_t tcps_pcbcachemiss;
528 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
529 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
530 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
531 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
532 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
533 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
534 	uint64_t tcps_persistdrop;	/* timeout in persist state */
535 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
536 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
537 	uint64_t tcps_listendrop;	/* listen queue overflows */
538 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
539 
540 	uint64_t tcps_sc_added;		/* entry added to syncache */
541 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
542 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
543 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
544 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
545 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
546 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
547 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
548 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
549 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
550 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
551 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
552 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
553 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
554 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
555 
556 	uint64_t tcps_hc_added;		/* entry added to hostcache */
557 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
558 
559 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
560 
561 	/* SACK related stats */
562 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
563 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
564 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
565 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
566 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
567 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
568 
569 	/* ECN related stats */
570 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
571 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
572 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
573 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
574 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
575 
576 	/* TCP_SIGNATURE related stats */
577 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
578 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
579 	uint64_t tcps_sig_err_buildsig;	/* Failed to make signature */
580 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
581 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
582 
583 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
584 };
585 
586 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
587 
588 #ifdef _KERNEL
589 #define	TI_UNLOCKED	1
590 #define	TI_RLOCKED	2
591 #include <sys/counter.h>
592 
593 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
594 /*
595  * In-kernel consumers can use these accessor macros directly to update
596  * stats.
597  */
598 #define	TCPSTAT_ADD(name, val)	\
599     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
600 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
601 
602 /*
603  * Kernel module consumers must use this accessor macro.
604  */
605 void	kmod_tcpstat_inc(int statnum);
606 #define	KMOD_TCPSTAT_INC(name)						\
607     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
608 
609 /*
610  * Running TCP connection count by state.
611  */
612 VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
613 #define	V_tcps_states	VNET(tcps_states)
614 #define	TCPSTATES_INC(state)	counter_u64_add(V_tcps_states[state], 1)
615 #define	TCPSTATES_DEC(state)	counter_u64_add(V_tcps_states[state], -1)
616 
617 /*
618  * TCP specific helper hook point identifiers.
619  */
620 #define	HHOOK_TCP_EST_IN		0
621 #define	HHOOK_TCP_EST_OUT		1
622 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
623 
624 struct tcp_hhook_data {
625 	struct tcpcb	*tp;
626 	struct tcphdr	*th;
627 	struct tcpopt	*to;
628 	uint32_t	len;
629 	int		tso;
630 	tcp_seq		curack;
631 };
632 #endif
633 
634 /*
635  * TCB structure exported to user-land via sysctl(3).
636  *
637  * Fields prefixed with "xt_" are unique to the export structure, and fields
638  * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
639  *
640  * Legend:
641  * (s) - used by userland utilities in src
642  * (p) - used by utilities in ports
643  * (3) - is known to be used by third party software not in ports
644  * (n) - no known usage
645  *
646  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
647  * included.  Not all of our clients do.
648  */
649 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
650 struct xtcpcb {
651 	size_t		xt_len;		/* length of this structure */
652 	struct xinpcb	xt_inp;
653 	char		xt_stack[TCP_FUNCTION_NAME_LEN_MAX];	/* (n) */
654 	int64_t		spare64[8];
655 	int32_t		t_state;		/* (s,p) */
656 	uint32_t	t_flags;		/* (s,p) */
657 	int32_t		t_sndzerowin;		/* (s) */
658 	int32_t		t_sndrexmitpack;	/* (s) */
659 	int32_t		t_rcvoopack;		/* (s) */
660 	int32_t		t_rcvtime;		/* (s) */
661 	int32_t		tt_rexmt;		/* (s) */
662 	int32_t		tt_persist;		/* (s) */
663 	int32_t		tt_keep;		/* (s) */
664 	int32_t		tt_2msl;		/* (s) */
665 	int32_t		tt_delack;		/* (s) */
666 	int32_t		spare32[32];
667 } __aligned(8);
668 #ifdef _KERNEL
669 void	tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
670 #endif
671 #endif
672 
673 /*
674  * Identifiers for TCP sysctl nodes
675  */
676 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
677 #define	TCPCTL_MSSDFLT		3	/* MSS default */
678 #define TCPCTL_STATS		4	/* statistics */
679 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
680 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
681 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
682 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
683 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
684 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
685 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
686 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
687 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
688 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
689 #define	TCPCTL_DROP		15	/* drop tcp connection */
690 #define	TCPCTL_STATES		16	/* connection counts by TCP state */
691 
692 #ifdef _KERNEL
693 #ifdef SYSCTL_DECL
694 SYSCTL_DECL(_net_inet_tcp);
695 SYSCTL_DECL(_net_inet_tcp_sack);
696 MALLOC_DECLARE(M_TCPLOG);
697 #endif
698 
699 VNET_DECLARE(struct inpcbhead, tcb);		/* queue of active tcpcb's */
700 VNET_DECLARE(struct inpcbinfo, tcbinfo);
701 extern	int tcp_log_in_vain;
702 VNET_DECLARE(int, tcp_mssdflt);	/* XXX */
703 VNET_DECLARE(int, tcp_minmss);
704 VNET_DECLARE(int, tcp_delack_enabled);
705 VNET_DECLARE(int, tcp_do_rfc3390);
706 VNET_DECLARE(int, tcp_initcwnd_segments);
707 VNET_DECLARE(int, tcp_sendspace);
708 VNET_DECLARE(int, tcp_recvspace);
709 VNET_DECLARE(int, path_mtu_discovery);
710 VNET_DECLARE(int, tcp_do_rfc3465);
711 VNET_DECLARE(int, tcp_abc_l_var);
712 #define	V_tcb			VNET(tcb)
713 #define	V_tcbinfo		VNET(tcbinfo)
714 #define	V_tcp_mssdflt		VNET(tcp_mssdflt)
715 #define	V_tcp_minmss		VNET(tcp_minmss)
716 #define	V_tcp_delack_enabled	VNET(tcp_delack_enabled)
717 #define	V_tcp_do_rfc3390	VNET(tcp_do_rfc3390)
718 #define	V_tcp_initcwnd_segments	VNET(tcp_initcwnd_segments)
719 #define	V_tcp_sendspace		VNET(tcp_sendspace)
720 #define	V_tcp_recvspace		VNET(tcp_recvspace)
721 #define	V_path_mtu_discovery	VNET(path_mtu_discovery)
722 #define	V_tcp_do_rfc3465	VNET(tcp_do_rfc3465)
723 #define	V_tcp_abc_l_var		VNET(tcp_abc_l_var)
724 
725 VNET_DECLARE(int, tcp_do_sack);			/* SACK enabled/disabled */
726 VNET_DECLARE(int, tcp_sc_rst_sock_fail);	/* RST on sock alloc failure */
727 #define	V_tcp_do_sack		VNET(tcp_do_sack)
728 #define	V_tcp_sc_rst_sock_fail	VNET(tcp_sc_rst_sock_fail)
729 
730 VNET_DECLARE(int, tcp_do_ecn);			/* TCP ECN enabled/disabled */
731 VNET_DECLARE(int, tcp_ecn_maxretries);
732 #define	V_tcp_do_ecn		VNET(tcp_do_ecn)
733 #define	V_tcp_ecn_maxretries	VNET(tcp_ecn_maxretries)
734 
735 #ifdef TCP_HHOOK
736 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
737 #define	V_tcp_hhh		VNET(tcp_hhh)
738 #endif
739 
740 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
741 #define V_tcp_do_rfc6675_pipe	VNET(tcp_do_rfc6675_pipe)
742 
743 int	 tcp_addoptions(struct tcpopt *, u_char *);
744 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
745 struct tcpcb *
746 	 tcp_close(struct tcpcb *);
747 void	 tcp_discardcb(struct tcpcb *);
748 void	 tcp_twstart(struct tcpcb *);
749 void	 tcp_twclose(struct tcptw *, int);
750 void	 tcp_ctlinput(int, struct sockaddr *, void *);
751 int	 tcp_ctloutput(struct socket *, struct sockopt *);
752 struct tcpcb *
753 	 tcp_drop(struct tcpcb *, int);
754 void	 tcp_drain(void);
755 void	 tcp_init(void);
756 void	 tcp_fini(void *);
757 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
758 	    const void *);
759 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
760 	    const void *);
761 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
762 void	 tcp_reass_global_init(void);
763 void	 tcp_reass_flush(struct tcpcb *);
764 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
765 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
766 		     struct tcpcb *, int, int);
767 void	tcp_pulloutofband(struct socket *,
768 		     struct tcphdr *, struct mbuf *, int);
769 void	tcp_xmit_timer(struct tcpcb *, int);
770 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
771 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
772 			    uint16_t nsegs, uint16_t type);
773 void 	cc_conn_init(struct tcpcb *tp);
774 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
775 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
776 #ifdef TCP_HHOOK
777 void	hhook_run_tcp_est_in(struct tcpcb *tp,
778 			    struct tcphdr *th, struct tcpopt *to);
779 #endif
780 
781 int	 tcp_input(struct mbuf **, int *, int);
782 int	 tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
783 	    struct tcpcb *, int);
784 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
785 			struct socket *, struct tcpcb *, int, int, uint8_t,
786 			int);
787 
788 int register_tcp_functions(struct tcp_function_block *blk, int wait);
789 int register_tcp_functions_as_names(struct tcp_function_block *blk,
790     int wait, const char *names[], int *num_names);
791 int register_tcp_functions_as_name(struct tcp_function_block *blk,
792     const char *name, int wait);
793 int deregister_tcp_functions(struct tcp_function_block *blk);
794 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
795 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
796 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
797 
798 uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
799 uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
800 u_int	 tcp_maxseg(const struct tcpcb *);
801 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
802 	    struct tcp_ifcap *);
803 void	 tcp_mss(struct tcpcb *, int);
804 int	 tcp_mssopt(struct in_conninfo *);
805 struct inpcb *
806 	 tcp_drop_syn_sent(struct inpcb *, int);
807 struct tcpcb *
808 	 tcp_newtcpcb(struct inpcb *);
809 int	 tcp_output(struct tcpcb *);
810 void	 tcp_state_change(struct tcpcb *, int);
811 void	 tcp_respond(struct tcpcb *, void *,
812 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
813 void	 tcp_tw_init(void);
814 #ifdef VIMAGE
815 void	 tcp_tw_destroy(void);
816 #endif
817 void	 tcp_tw_zone_change(void);
818 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
819 	    struct mbuf *, int);
820 void	 tcp_setpersist(struct tcpcb *);
821 void	 tcp_slowtimo(void);
822 struct tcptemp *
823 	 tcpip_maketemplate(struct inpcb *);
824 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
825 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
826 int	 tcp_timer_active(struct tcpcb *, uint32_t);
827 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
828 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
829 /*
830  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
831  */
832 void	 tcp_hc_init(void);
833 #ifdef VIMAGE
834 void	 tcp_hc_destroy(void);
835 #endif
836 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
837 uint32_t tcp_hc_getmtu(struct in_conninfo *);
838 void	 tcp_hc_updatemtu(struct in_conninfo *, uint32_t);
839 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
840 
841 extern	struct pr_usrreqs tcp_usrreqs;
842 tcp_seq tcp_new_isn(struct tcpcb *);
843 
844 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
845 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
846 void	 tcp_clean_sackreport(struct tcpcb *tp);
847 void	 tcp_sack_adjust(struct tcpcb *tp);
848 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
849 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
850 void	 tcp_free_sackholes(struct tcpcb *tp);
851 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
852 int	 tcp_compute_pipe(struct tcpcb *);
853 
854 static inline void
855 tcp_fields_to_host(struct tcphdr *th)
856 {
857 
858 	th->th_seq = ntohl(th->th_seq);
859 	th->th_ack = ntohl(th->th_ack);
860 	th->th_win = ntohs(th->th_win);
861 	th->th_urp = ntohs(th->th_urp);
862 }
863 
864 static inline void
865 tcp_fields_to_net(struct tcphdr *th)
866 {
867 
868 	th->th_seq = htonl(th->th_seq);
869 	th->th_ack = htonl(th->th_ack);
870 	th->th_win = htons(th->th_win);
871 	th->th_urp = htons(th->th_urp);
872 }
873 #endif /* _KERNEL */
874 
875 #endif /* _NETINET_TCP_VAR_H_ */
876