xref: /freebsd/sys/netinet/tcp_var.h (revision a0ee8cc636cd5c2374ec44ca71226564ea0bca95)
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  * 4. 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 
38 #ifdef _KERNEL
39 #include <net/vnet.h>
40 #include <sys/mbuf.h>
41 
42 /*
43  * Kernel variables for tcp.
44  */
45 VNET_DECLARE(int, tcp_do_rfc1323);
46 #define	V_tcp_do_rfc1323	VNET(tcp_do_rfc1323)
47 
48 #endif /* _KERNEL */
49 
50 /* TCP segment queue entry */
51 struct tseg_qent {
52 	LIST_ENTRY(tseg_qent) tqe_q;
53 	int	tqe_len;		/* TCP segment data length */
54 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
55 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
56 };
57 LIST_HEAD(tsegqe_head, tseg_qent);
58 
59 struct sackblk {
60 	tcp_seq start;		/* start seq no. of sack block */
61 	tcp_seq end;		/* end seq no. */
62 };
63 
64 struct sackhole {
65 	tcp_seq start;		/* start seq no. of hole */
66 	tcp_seq end;		/* end seq no. */
67 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
68 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
69 };
70 
71 struct sackhint {
72 	struct sackhole	*nexthole;
73 	int		sack_bytes_rexmit;
74 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
75 
76 	int		ispare;		/* explicit pad for 64bit alignment */
77 	int             sacked_bytes;	/*
78 					 * Total sacked bytes reported by the
79 					 * receiver via sack option
80 					 */
81 	uint32_t	_pad1[1];	/* TBD */
82 	uint64_t	_pad[1];	/* TBD */
83 };
84 
85 struct tcptemp {
86 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
87 	struct	tcphdr tt_t;
88 };
89 
90 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
91 
92 /*
93  * TODO: We yet need to brave plowing in
94  * to tcp_input() and the pru_usrreq() block.
95  * Right now these go to the old standards which
96  * are somewhat ok, but in the long term may
97  * need to be changed. If we do tackle tcp_input()
98  * then we need to get rid of the tcp_do_segment()
99  * function below.
100  */
101 /* Flags for tcp functions */
102 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
103 struct tcpcb;
104 struct inpcb;
105 struct sockopt;
106 struct socket;
107 
108 struct tcp_function_block {
109 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
110 	int	(*tfb_tcp_output)(struct tcpcb *);
111 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
112 			    struct socket *, struct tcpcb *,
113 			    int, int, uint8_t,
114 			    int);
115 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
116 			    struct inpcb *inp, struct tcpcb *tp);
117 	/* Optional memory allocation/free routine */
118 	void	(*tfb_tcp_fb_init)(struct tcpcb *);
119 	void	(*tfb_tcp_fb_fini)(struct tcpcb *);
120 	/* Optional timers, must define all if you define one */
121 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
122 	int	(*tfb_tcp_timers_left)(struct tcpcb *);
123 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
124 			    uint32_t, u_int);
125 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
126 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
127 	volatile uint32_t tfb_refcnt;
128 	uint32_t  tfb_flags;
129 };
130 
131 struct tcp_function {
132 	TAILQ_ENTRY(tcp_function) tf_next;
133 	struct tcp_function_block *tf_fb;
134 };
135 
136 TAILQ_HEAD(tcp_funchead, tcp_function);
137 
138 /*
139  * Tcp control block, one per tcp; fields:
140  * Organized for 16 byte cacheline efficiency.
141  */
142 struct tcpcb {
143 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
144 	void	*t_pspare[2];		/* new reassembly queue */
145 	int	t_segqlen;		/* segment reassembly queue length */
146 	int	t_dupacks;		/* consecutive dup acks recd */
147 
148 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
149 
150 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
151 	int	t_state;		/* state of this connection */
152 	u_int	t_flags;
153 
154 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
155 
156 	tcp_seq	snd_una;		/* sent but unacknowledged */
157 	tcp_seq	snd_max;		/* highest sequence number sent;
158 					 * used to recognize retransmits
159 					 */
160 	tcp_seq	snd_nxt;		/* send next */
161 	tcp_seq	snd_up;			/* send urgent pointer */
162 
163 	tcp_seq	snd_wl1;		/* window update seg seq number */
164 	tcp_seq	snd_wl2;		/* window update seg ack number */
165 	tcp_seq	iss;			/* initial send sequence number */
166 	tcp_seq	irs;			/* initial receive sequence number */
167 
168 	tcp_seq	rcv_nxt;		/* receive next */
169 	tcp_seq	rcv_adv;		/* advertised window */
170 	u_long	rcv_wnd;		/* receive window */
171 	tcp_seq	rcv_up;			/* receive urgent pointer */
172 
173 	u_long	snd_wnd;		/* send window */
174 	u_long	snd_cwnd;		/* congestion-controlled window */
175 	u_long	snd_spare1;		/* unused */
176 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
177 					 * for slow start exponential to
178 					 * linear switch
179 					 */
180 	u_long	snd_spare2;		/* unused */
181 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
182 
183 	u_int	t_rcvtime;		/* inactivity time */
184 	u_int	t_starttime;		/* time connection was established */
185 	u_int	t_rtttime;		/* RTT measurement start time */
186 	tcp_seq	t_rtseq;		/* sequence number being timed */
187 
188 	u_int	t_bw_spare1;		/* unused */
189 	tcp_seq	t_bw_spare2;		/* unused */
190 
191 	int	t_rxtcur;		/* current retransmit value (ticks) */
192 	u_int	t_maxseg;		/* maximum segment size */
193 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
194 	int	t_srtt;			/* smoothed round-trip time */
195 	int	t_rttvar;		/* variance in round-trip time */
196 
197 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
198 	u_int	t_rttmin;		/* minimum rtt allowed */
199 	u_int	t_rttbest;		/* best rtt we've seen */
200 	u_long	t_rttupdated;		/* number of times rtt sampled */
201 	u_long	max_sndwnd;		/* largest window peer has offered */
202 
203 	int	t_softerror;		/* possible error not yet reported */
204 /* out-of-band data */
205 	char	t_oobflags;		/* have some */
206 	char	t_iobc;			/* input character */
207 /* RFC 1323 variables */
208 	u_char	snd_scale;		/* window scaling for send window */
209 	u_char	rcv_scale;		/* window scaling for recv window */
210 	u_char	request_r_scale;	/* pending window scaling */
211 	u_int32_t  ts_recent;		/* timestamp echo data */
212 	u_int	ts_recent_age;		/* when last updated */
213 	u_int32_t  ts_offset;		/* our timestamp offset */
214 
215 	tcp_seq	last_ack_sent;
216 /* experimental */
217 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
218 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
219 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
220 	int	t_sndzerowin;		/* zero-window updates sent */
221 	u_int	t_badrxtwin;		/* window for retransmit recovery */
222 	u_char	snd_limited;		/* segments limited transmitted */
223 /* SACK related state */
224 	int	snd_numholes;		/* number of holes seen by sender */
225 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
226 					/* SACK scoreboard (sorted) */
227 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
228 	int	rcv_numsacks;		/* # distinct sack blks present */
229 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
230 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
231 					   episode starts at this seq number */
232 	struct sackhint	sackhint;	/* SACK scoreboard hint */
233 	int	t_rttlow;		/* smallest observerved RTT */
234 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
235 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
236 	struct toedev	*tod;		/* toedev handling this connection */
237 	int	t_sndrexmitpack;	/* retransmit packets sent */
238 	int	t_rcvoopack;		/* out-of-order packets received */
239 	void	*t_toe;			/* TOE pcb pointer */
240 	int	t_bytes_acked;		/* # bytes acked during current RTT */
241 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
242 	struct cc_var	*ccv;		/* congestion control specific vars */
243 	struct osd	*osd;		/* storage for Khelp module data */
244 
245 	u_int	t_keepinit;		/* time to establish connection */
246 	u_int	t_keepidle;		/* time before keepalive probes begin */
247 	u_int	t_keepintvl;		/* interval between keepalives */
248 	u_int	t_keepcnt;		/* number of keepalives before close */
249 
250 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
251 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
252 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
253 	u_int	t_flags2;		/* More tcpcb flags storage */
254 #if defined(_KERNEL) && defined(TCP_RFC7413)
255 	uint32_t t_ispare[6];		/* 5 UTO, 1 TBD */
256 	uint64_t t_tfo_cookie;		/* TCP Fast Open cookie */
257 #else
258 	uint32_t t_ispare[8];		/* 5 UTO, 3 TBD */
259 #endif
260 	struct tcp_function_block *t_fb;/* TCP function call block */
261 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
262 #if defined(_KERNEL) && defined(TCP_RFC7413)
263 	unsigned int *t_tfo_pending;	/* TCP Fast Open pending counter */
264 	void	*t_pspare2[1];		/* 1 TCP_SIGNATURE */
265 #else
266 	void	*t_pspare2[2];		/* 1 TCP_SIGNATURE, 1 TBD */
267 #endif
268 #if defined(_KERNEL) && defined(TCPPCAP)
269 	struct mbufq t_inpkts;		/* List of saved input packets. */
270 	struct mbufq t_outpkts;		/* List of saved output packets. */
271 #ifdef _LP64
272 	uint64_t _pad[0];		/* all used! */
273 #else
274 	uint64_t _pad[2];		/* 2 are available */
275 #endif /* _LP64 */
276 #else
277 	uint64_t _pad[6];
278 #endif /* defined(_KERNEL) && defined(TCPPCAP) */
279 };
280 
281 /*
282  * Flags and utility macros for the t_flags field.
283  */
284 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
285 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
286 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
287 #define	TF_NOOPT	0x000008	/* don't use tcp options */
288 #define	TF_SENTFIN	0x000010	/* have sent FIN */
289 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
290 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
291 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
292 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
293 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
294 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
295 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
296 #define	TF_NOPUSH	0x001000	/* don't push */
297 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
298 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
299 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
300 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
301 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
302 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
303 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
304 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
305 #define	TF_FORCEDATA	0x800000	/* force out a byte */
306 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
307 #define	TF_TOE		0x2000000	/* this connection is offloaded */
308 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
309 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
310 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
311 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
312 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
313 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
314 
315 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
316 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
317 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
318 
319 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
320 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
321 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
322 
323 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
324 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
325 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
326 
327 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
328 
329 /*
330  * Flags for the t_oobflags field.
331  */
332 #define	TCPOOB_HAVEDATA	0x01
333 #define	TCPOOB_HADDATA	0x02
334 
335 #ifdef TCP_SIGNATURE
336 /*
337  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
338  * for SADB verification and lookup.
339  */
340 #define	TCP_SIGLEN	16	/* length of computed digest in bytes */
341 #define	TCP_KEYLEN_MIN	1	/* minimum length of TCP-MD5 key */
342 #define	TCP_KEYLEN_MAX	80	/* maximum length of TCP-MD5 key */
343 /*
344  * Only a single SA per host may be specified at this time. An SPI is
345  * needed in order for the KEY_ALLOCSA() lookup to work.
346  */
347 #define	TCP_SIG_SPI	0x1000
348 #endif /* TCP_SIGNATURE */
349 
350 /*
351  * Flags for PLPMTU handling, t_flags2
352  */
353 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
354 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
355 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
356 
357 /*
358  * Structure to hold TCP options that are only used during segment
359  * processing (in tcp_input), but not held in the tcpcb.
360  * It's basically used to reduce the number of parameters
361  * to tcp_dooptions and tcp_addoptions.
362  * The binary order of the to_flags is relevant for packing of the
363  * options in tcp_addoptions.
364  */
365 struct tcpopt {
366 	u_int64_t	to_flags;	/* which options are present */
367 #define	TOF_MSS		0x0001		/* maximum segment size */
368 #define	TOF_SCALE	0x0002		/* window scaling */
369 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
370 #define	TOF_TS		0x0010		/* timestamp */
371 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
372 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
373 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
374 #define	TOF_MAXOPT	0x0200
375 	u_int32_t	to_tsval;	/* new timestamp */
376 	u_int32_t	to_tsecr;	/* reflected timestamp */
377 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
378 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
379 	u_char		*to_tfo_cookie; /* pointer to the TFO cookie */
380 	u_int16_t	to_mss;		/* maximum segment size */
381 	u_int8_t	to_wscale;	/* window scaling */
382 	u_int8_t	to_nsacks;	/* number of SACK blocks */
383 	u_int8_t	to_tfo_len;	/* TFO cookie length */
384 	u_int32_t	to_spare;	/* UTO */
385 };
386 
387 /*
388  * Flags for tcp_dooptions.
389  */
390 #define	TO_SYN		0x01		/* parse SYN-only options */
391 
392 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
393 	u_long	rmx_mtu;	/* MTU for this path */
394 	u_long	rmx_ssthresh;	/* outbound gateway buffer limit */
395 	u_long	rmx_rtt;	/* estimated round trip time */
396 	u_long	rmx_rttvar;	/* estimated rtt variance */
397 	u_long	rmx_cwnd;	/* congestion window */
398 	u_long	rmx_sendpipe;   /* outbound delay-bandwidth product */
399 	u_long	rmx_recvpipe;   /* inbound delay-bandwidth product */
400 };
401 
402 /*
403  * Used by tcp_maxmtu() to communicate interface specific features
404  * and limits at the time of connection setup.
405  */
406 struct tcp_ifcap {
407 	int	ifcap;
408 	u_int	tsomax;
409 	u_int	tsomaxsegcount;
410 	u_int	tsomaxsegsize;
411 };
412 
413 #ifndef _NETINET_IN_PCB_H_
414 struct in_conninfo;
415 #endif /* _NETINET_IN_PCB_H_ */
416 
417 struct tcptw {
418 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
419 	tcp_seq		snd_nxt;
420 	tcp_seq		rcv_nxt;
421 	tcp_seq		iss;
422 	tcp_seq		irs;
423 	u_short		last_win;	/* cached window value */
424 	u_short		tw_so_options;	/* copy of so_options */
425 	struct ucred	*tw_cred;	/* user credentials */
426 	u_int32_t	t_recent;
427 	u_int32_t	ts_offset;	/* our timestamp offset */
428 	u_int		t_starttime;
429 	int		tw_time;
430 	TAILQ_ENTRY(tcptw) tw_2msl;
431 	void		*tw_pspare;	/* TCP_SIGNATURE */
432 	u_int		*tw_spare;	/* TCP_SIGNATURE */
433 };
434 
435 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
436 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
437 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
438 
439 /*
440  * The smoothed round-trip time and estimated variance
441  * are stored as fixed point numbers scaled by the values below.
442  * For convenience, these scales are also used in smoothing the average
443  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
444  * With these scales, srtt has 3 bits to the right of the binary point,
445  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
446  * binary point, and is smoothed with an ALPHA of 0.75.
447  */
448 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
449 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
450 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
451 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
452 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
453 
454 /*
455  * The initial retransmission should happen at rtt + 4 * rttvar.
456  * Because of the way we do the smoothing, srtt and rttvar
457  * will each average +1/2 tick of bias.  When we compute
458  * the retransmit timer, we want 1/2 tick of rounding and
459  * 1 extra tick because of +-1/2 tick uncertainty in the
460  * firing of the timer.  The bias will give us exactly the
461  * 1.5 tick we need.  But, because the bias is
462  * statistical, we have to test that we don't drop below
463  * the minimum feasible timer (which is 2 ticks).
464  * This version of the macro adapted from a paper by Lawrence
465  * Brakmo and Larry Peterson which outlines a problem caused
466  * by insufficient precision in the original implementation,
467  * which results in inappropriately large RTO values for very
468  * fast networks.
469  */
470 #define	TCP_REXMTVAL(tp) \
471 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
472 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
473 
474 /*
475  * TCP statistics.
476  * Many of these should be kept per connection,
477  * but that's inconvenient at the moment.
478  */
479 struct	tcpstat {
480 	uint64_t tcps_connattempt;	/* connections initiated */
481 	uint64_t tcps_accepts;		/* connections accepted */
482 	uint64_t tcps_connects;		/* connections established */
483 	uint64_t tcps_drops;		/* connections dropped */
484 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
485 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
486 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
487 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
488 	uint64_t tcps_rttupdated;	/* times we succeeded */
489 	uint64_t tcps_delack;		/* delayed acks sent */
490 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
491 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
492 	uint64_t tcps_persisttimeo;	/* persist timeouts */
493 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
494 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
495 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
496 
497 	uint64_t tcps_sndtotal;		/* total packets sent */
498 	uint64_t tcps_sndpack;		/* data packets sent */
499 	uint64_t tcps_sndbyte;		/* data bytes sent */
500 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
501 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
502 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
503 	uint64_t tcps_sndacks;		/* ack-only packets sent */
504 	uint64_t tcps_sndprobe;		/* window probes sent */
505 	uint64_t tcps_sndurg;		/* packets sent with URG only */
506 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
507 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
508 
509 	uint64_t tcps_rcvtotal;		/* total packets received */
510 	uint64_t tcps_rcvpack;		/* packets received in sequence */
511 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
512 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
513 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
514 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
515 	uint64_t tcps_rcvshort;		/* packets received too short */
516 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
517 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
518 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
519 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
520 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
521 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
522 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
523 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
524 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
525 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
526 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
527 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
528 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
529 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
530 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
531 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
532 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
533 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
534 	uint64_t tcps_pcbcachemiss;
535 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
536 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
537 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
538 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
539 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
540 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
541 	uint64_t tcps_persistdrop;	/* timeout in persist state */
542 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
543 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
544 	uint64_t tcps_listendrop;	/* listen queue overflows */
545 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
546 
547 	uint64_t tcps_sc_added;		/* entry added to syncache */
548 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
549 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
550 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
551 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
552 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
553 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
554 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
555 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
556 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
557 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
558 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
559 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
560 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
561 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
562 
563 	uint64_t tcps_hc_added;		/* entry added to hostcache */
564 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
565 
566 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
567 
568 	/* SACK related stats */
569 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
570 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
571 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
572 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
573 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
574 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
575 
576 	/* ECN related stats */
577 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
578 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
579 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
580 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
581 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
582 
583 	/* TCP_SIGNATURE related stats */
584 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
585 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
586 	uint64_t tcps_sig_err_buildsig;	/* Mismatching signature received */
587 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
588 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
589 
590 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
591 };
592 
593 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
594 
595 #ifdef _KERNEL
596 #define	TI_UNLOCKED	1
597 #define	TI_RLOCKED	2
598 #include <sys/counter.h>
599 
600 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
601 /*
602  * In-kernel consumers can use these accessor macros directly to update
603  * stats.
604  */
605 #define	TCPSTAT_ADD(name, val)	\
606     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
607 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
608 
609 /*
610  * Kernel module consumers must use this accessor macro.
611  */
612 void	kmod_tcpstat_inc(int statnum);
613 #define	KMOD_TCPSTAT_INC(name)						\
614     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
615 
616 /*
617  * TCP specific helper hook point identifiers.
618  */
619 #define	HHOOK_TCP_EST_IN		0
620 #define	HHOOK_TCP_EST_OUT		1
621 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
622 
623 struct tcp_hhook_data {
624 	struct tcpcb	*tp;
625 	struct tcphdr	*th;
626 	struct tcpopt	*to;
627 	long		len;
628 	int		tso;
629 	tcp_seq		curack;
630 };
631 #endif
632 
633 /*
634  * TCB structure exported to user-land via sysctl(3).
635  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
636  * included.  Not all of our clients do.
637  */
638 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
639 struct xtcp_timer {
640 	int tt_rexmt;	/* retransmit timer */
641 	int tt_persist;	/* retransmit persistence */
642 	int tt_keep;	/* keepalive */
643 	int tt_2msl;	/* 2*msl TIME_WAIT timer */
644 	int tt_delack;	/* delayed ACK timer */
645 	int t_rcvtime;	/* Time since last packet received */
646 };
647 struct	xtcpcb {
648 	size_t	xt_len;
649 	struct	inpcb	xt_inp;
650 	struct	tcpcb	xt_tp;
651 	struct	xsocket	xt_socket;
652 	struct	xtcp_timer xt_timer;
653 	u_quad_t	xt_alignment_hack;
654 };
655 #endif
656 
657 /*
658  * Identifiers for TCP sysctl nodes
659  */
660 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
661 #define	TCPCTL_MSSDFLT		3	/* MSS default */
662 #define TCPCTL_STATS		4	/* statistics (read-only) */
663 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
664 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
665 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
666 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
667 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
668 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
669 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
670 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
671 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
672 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
673 #define	TCPCTL_DROP		15	/* drop tcp connection */
674 
675 #ifdef _KERNEL
676 #ifdef SYSCTL_DECL
677 SYSCTL_DECL(_net_inet_tcp);
678 SYSCTL_DECL(_net_inet_tcp_sack);
679 MALLOC_DECLARE(M_TCPLOG);
680 #endif
681 
682 VNET_DECLARE(struct inpcbhead, tcb);		/* queue of active tcpcb's */
683 VNET_DECLARE(struct inpcbinfo, tcbinfo);
684 extern	int tcp_log_in_vain;
685 VNET_DECLARE(int, tcp_mssdflt);	/* XXX */
686 VNET_DECLARE(int, tcp_minmss);
687 VNET_DECLARE(int, tcp_delack_enabled);
688 VNET_DECLARE(int, tcp_do_rfc3390);
689 VNET_DECLARE(int, tcp_initcwnd_segments);
690 VNET_DECLARE(int, tcp_sendspace);
691 VNET_DECLARE(int, tcp_recvspace);
692 VNET_DECLARE(int, path_mtu_discovery);
693 VNET_DECLARE(int, tcp_do_rfc3465);
694 VNET_DECLARE(int, tcp_abc_l_var);
695 #define	V_tcb			VNET(tcb)
696 #define	V_tcbinfo		VNET(tcbinfo)
697 #define	V_tcp_mssdflt		VNET(tcp_mssdflt)
698 #define	V_tcp_minmss		VNET(tcp_minmss)
699 #define	V_tcp_delack_enabled	VNET(tcp_delack_enabled)
700 #define	V_tcp_do_rfc3390	VNET(tcp_do_rfc3390)
701 #define	V_tcp_initcwnd_segments	VNET(tcp_initcwnd_segments)
702 #define	V_tcp_sendspace		VNET(tcp_sendspace)
703 #define	V_tcp_recvspace		VNET(tcp_recvspace)
704 #define	V_path_mtu_discovery	VNET(path_mtu_discovery)
705 #define	V_tcp_do_rfc3465	VNET(tcp_do_rfc3465)
706 #define	V_tcp_abc_l_var		VNET(tcp_abc_l_var)
707 
708 VNET_DECLARE(int, tcp_do_sack);			/* SACK enabled/disabled */
709 VNET_DECLARE(int, tcp_sc_rst_sock_fail);	/* RST on sock alloc failure */
710 #define	V_tcp_do_sack		VNET(tcp_do_sack)
711 #define	V_tcp_sc_rst_sock_fail	VNET(tcp_sc_rst_sock_fail)
712 
713 VNET_DECLARE(int, tcp_do_ecn);			/* TCP ECN enabled/disabled */
714 VNET_DECLARE(int, tcp_ecn_maxretries);
715 #define	V_tcp_do_ecn		VNET(tcp_do_ecn)
716 #define	V_tcp_ecn_maxretries	VNET(tcp_ecn_maxretries)
717 
718 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
719 #define	V_tcp_hhh		VNET(tcp_hhh)
720 
721 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
722 #define V_tcp_do_rfc6675_pipe	VNET(tcp_do_rfc6675_pipe)
723 
724 int	 tcp_addoptions(struct tcpopt *, u_char *);
725 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
726 struct tcpcb *
727 	 tcp_close(struct tcpcb *);
728 void	 tcp_discardcb(struct tcpcb *);
729 void	 tcp_twstart(struct tcpcb *);
730 void	 tcp_twclose(struct tcptw *, int);
731 void	 tcp_ctlinput(int, struct sockaddr *, void *);
732 int	 tcp_ctloutput(struct socket *, struct sockopt *);
733 struct tcpcb *
734 	 tcp_drop(struct tcpcb *, int);
735 void	 tcp_drain(void);
736 void	 tcp_init(void);
737 #ifdef VIMAGE
738 void	 tcp_destroy(void);
739 #endif
740 void	 tcp_fini(void *);
741 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
742 	    const void *);
743 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
744 	    const void *);
745 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
746 void	 tcp_reass_global_init(void);
747 void	 tcp_reass_flush(struct tcpcb *);
748 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
749 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
750 		     struct tcpcb *, int, int);
751 void	tcp_pulloutofband(struct socket *,
752 		     struct tcphdr *, struct mbuf *, int);
753 void	tcp_xmit_timer(struct tcpcb *, int);
754 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
755 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
756 			    uint16_t type);
757 void 	cc_conn_init(struct tcpcb *tp);
758 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
759 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
760 void	hhook_run_tcp_est_in(struct tcpcb *tp,
761 			    struct tcphdr *th, struct tcpopt *to);
762 
763 int	 tcp_input(struct mbuf **, int *, int);
764 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
765 			struct socket *, struct tcpcb *, int, int, uint8_t,
766 			int);
767 
768 int register_tcp_functions(struct tcp_function_block *blk, int wait);
769 int deregister_tcp_functions(struct tcp_function_block *blk);
770 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
771 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
772 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
773 
774 u_long	 tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
775 u_long	 tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
776 u_int	 tcp_maxseg(const struct tcpcb *);
777 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
778 	    struct tcp_ifcap *);
779 void	 tcp_mss(struct tcpcb *, int);
780 int	 tcp_mssopt(struct in_conninfo *);
781 struct inpcb *
782 	 tcp_drop_syn_sent(struct inpcb *, int);
783 struct tcpcb *
784 	 tcp_newtcpcb(struct inpcb *);
785 int	 tcp_output(struct tcpcb *);
786 void	 tcp_state_change(struct tcpcb *, int);
787 void	 tcp_respond(struct tcpcb *, void *,
788 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
789 void	 tcp_tw_init(void);
790 #ifdef VIMAGE
791 void	 tcp_tw_destroy(void);
792 #endif
793 void	 tcp_tw_zone_change(void);
794 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
795 	    struct mbuf *, int);
796 void	 tcp_setpersist(struct tcpcb *);
797 #ifdef TCP_SIGNATURE
798 struct secasvar;
799 struct secasvar *tcp_get_sav(struct mbuf *, u_int);
800 int	 tcp_signature_do_compute(struct mbuf *, int, int, u_char *,
801 	    struct secasvar *);
802 int	 tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int);
803 int	 tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *,
804 	    struct tcphdr *, u_int);
805 int	tcp_signature_check(struct mbuf *m, int off0, int tlen, int optlen,
806 	    struct tcpopt *to, struct tcphdr *th, u_int tcpbflag);
807 #endif
808 void	 tcp_slowtimo(void);
809 struct tcptemp *
810 	 tcpip_maketemplate(struct inpcb *);
811 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
812 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
813 int	 tcp_timer_active(struct tcpcb *, uint32_t);
814 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
815 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
816 /*
817  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
818  */
819 void	 tcp_hc_init(void);
820 #ifdef VIMAGE
821 void	 tcp_hc_destroy(void);
822 #endif
823 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
824 u_long	 tcp_hc_getmtu(struct in_conninfo *);
825 void	 tcp_hc_updatemtu(struct in_conninfo *, u_long);
826 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
827 
828 extern	struct pr_usrreqs tcp_usrreqs;
829 tcp_seq tcp_new_isn(struct tcpcb *);
830 
831 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
832 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
833 void	 tcp_clean_sackreport(struct tcpcb *tp);
834 void	 tcp_sack_adjust(struct tcpcb *tp);
835 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
836 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
837 void	 tcp_free_sackholes(struct tcpcb *tp);
838 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
839 u_long	 tcp_seq_subtract(u_long, u_long );
840 int	 tcp_compute_pipe(struct tcpcb *);
841 
842 static inline void
843 tcp_fields_to_host(struct tcphdr *th)
844 {
845 
846 	th->th_seq = ntohl(th->th_seq);
847 	th->th_ack = ntohl(th->th_ack);
848 	th->th_win = ntohs(th->th_win);
849 	th->th_urp = ntohs(th->th_urp);
850 }
851 
852 #ifdef TCP_SIGNATURE
853 static inline void
854 tcp_fields_to_net(struct tcphdr *th)
855 {
856 
857 	th->th_seq = htonl(th->th_seq);
858 	th->th_ack = htonl(th->th_ack);
859 	th->th_win = htons(th->th_win);
860 	th->th_urp = htons(th->th_urp);
861 }
862 #endif
863 #endif /* _KERNEL */
864 
865 #endif /* _NETINET_TCP_VAR_H_ */
866