xref: /freebsd/sys/netinet/tcp_var.h (revision 7661de35d15f582ab33e3bd6b8d909601557e436)
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 
41 /*
42  * Kernel variables for tcp.
43  */
44 VNET_DECLARE(int, tcp_do_rfc1323);
45 #define	V_tcp_do_rfc1323	VNET(tcp_do_rfc1323)
46 
47 #endif /* _KERNEL */
48 
49 /* TCP segment queue entry */
50 struct tseg_qent {
51 	LIST_ENTRY(tseg_qent) tqe_q;
52 	int	tqe_len;		/* TCP segment data length */
53 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
54 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
55 };
56 LIST_HEAD(tsegqe_head, tseg_qent);
57 
58 struct sackblk {
59 	tcp_seq start;		/* start seq no. of sack block */
60 	tcp_seq end;		/* end seq no. */
61 };
62 
63 struct sackhole {
64 	tcp_seq start;		/* start seq no. of hole */
65 	tcp_seq end;		/* end seq no. */
66 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
67 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
68 };
69 
70 struct sackhint {
71 	struct sackhole	*nexthole;
72 	int		sack_bytes_rexmit;
73 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
74 
75 	int		ispare;		/* explicit pad for 64bit alignment */
76 	uint64_t	_pad[2];	/* 1 sacked_bytes, 1 TBD */
77 };
78 
79 struct tcptemp {
80 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
81 	struct	tcphdr tt_t;
82 };
83 
84 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
85 
86 /* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */
87 #ifdef INET6
88 #define ND6_HINT(tp)						\
89 do {								\
90 	if ((tp) && (tp)->t_inpcb &&				\
91 	    ((tp)->t_inpcb->inp_vflag & INP_IPV6) != 0)		\
92 		nd6_nud_hint(NULL, NULL, 0);			\
93 } while (0)
94 #else
95 #define ND6_HINT(tp)
96 #endif
97 
98 /*
99  * Tcp control block, one per tcp; fields:
100  * Organized for 16 byte cacheline efficiency.
101  */
102 struct tcpcb {
103 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
104 	void	*t_pspare[2];		/* new reassembly queue */
105 	int	t_segqlen;		/* segment reassembly queue length */
106 	int	t_dupacks;		/* consecutive dup acks recd */
107 
108 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
109 
110 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
111 	int	t_state;		/* state of this connection */
112 	u_int	t_flags;
113 
114 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
115 
116 	tcp_seq	snd_una;		/* send unacknowledged */
117 	tcp_seq	snd_max;		/* highest sequence number sent;
118 					 * used to recognize retransmits
119 					 */
120 	tcp_seq	snd_nxt;		/* send next */
121 	tcp_seq	snd_up;			/* send urgent pointer */
122 
123 	tcp_seq	snd_wl1;		/* window update seg seq number */
124 	tcp_seq	snd_wl2;		/* window update seg ack number */
125 	tcp_seq	iss;			/* initial send sequence number */
126 	tcp_seq	irs;			/* initial receive sequence number */
127 
128 	tcp_seq	rcv_nxt;		/* receive next */
129 	tcp_seq	rcv_adv;		/* advertised window */
130 	u_long	rcv_wnd;		/* receive window */
131 	tcp_seq	rcv_up;			/* receive urgent pointer */
132 
133 	u_long	snd_wnd;		/* send window */
134 	u_long	snd_cwnd;		/* congestion-controlled window */
135 	u_long	snd_spare1;		/* unused */
136 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
137 					 * for slow start exponential to
138 					 * linear switch
139 					 */
140 	u_long	snd_spare2;		/* unused */
141 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
142 
143 	u_int	t_maxopd;		/* mss plus options */
144 
145 	u_int	t_rcvtime;		/* inactivity time */
146 	u_int	t_starttime;		/* time connection was established */
147 	u_int	t_rtttime;		/* RTT measurement start time */
148 	tcp_seq	t_rtseq;		/* sequence number being timed */
149 
150 	u_int	t_bw_spare1;		/* unused */
151 	tcp_seq	t_bw_spare2;		/* unused */
152 
153 	int	t_rxtcur;		/* current retransmit value (ticks) */
154 	u_int	t_maxseg;		/* maximum segment size */
155 	int	t_srtt;			/* smoothed round-trip time */
156 	int	t_rttvar;		/* variance in round-trip time */
157 
158 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
159 	u_int	t_rttmin;		/* minimum rtt allowed */
160 	u_int	t_rttbest;		/* best rtt we've seen */
161 	u_long	t_rttupdated;		/* number of times rtt sampled */
162 	u_long	max_sndwnd;		/* largest window peer has offered */
163 
164 	int	t_softerror;		/* possible error not yet reported */
165 /* out-of-band data */
166 	char	t_oobflags;		/* have some */
167 	char	t_iobc;			/* input character */
168 /* RFC 1323 variables */
169 	u_char	snd_scale;		/* window scaling for send window */
170 	u_char	rcv_scale;		/* window scaling for recv window */
171 	u_char	request_r_scale;	/* pending window scaling */
172 	u_int32_t  ts_recent;		/* timestamp echo data */
173 	u_int	ts_recent_age;		/* when last updated */
174 	u_int32_t  ts_offset;		/* our timestamp offset */
175 
176 	tcp_seq	last_ack_sent;
177 /* experimental */
178 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
179 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
180 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
181 	int	t_sndzerowin;		/* zero-window updates sent */
182 	u_int	t_badrxtwin;		/* window for retransmit recovery */
183 	u_char	snd_limited;		/* segments limited transmitted */
184 /* SACK related state */
185 	int	snd_numholes;		/* number of holes seen by sender */
186 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
187 					/* SACK scoreboard (sorted) */
188 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
189 	int	rcv_numsacks;		/* # distinct sack blks present */
190 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
191 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
192 					   episode starts at this seq number */
193 	struct sackhint	sackhint;	/* SACK scoreboard hint */
194 	int	t_rttlow;		/* smallest observerved RTT */
195 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
196 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
197 	struct toedev	*tod;		/* toedev handling this connection */
198 	int	t_sndrexmitpack;	/* retransmit packets sent */
199 	int	t_rcvoopack;		/* out-of-order packets received */
200 	void	*t_toe;			/* TOE pcb pointer */
201 	int	t_bytes_acked;		/* # bytes acked during current RTT */
202 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
203 	struct cc_var	*ccv;		/* congestion control specific vars */
204 	struct osd	*osd;		/* storage for Khelp module data */
205 
206 	u_int	t_keepinit;		/* time to establish connection */
207 	u_int	t_keepidle;		/* time before keepalive probes begin */
208 	u_int	t_keepintvl;		/* interval between keepalives */
209 	u_int	t_keepcnt;		/* number of keepalives before close */
210 
211 	u_int	t_tsomax;		/* tso burst length limit */
212 
213 	uint32_t t_ispare[8];		/* 5 UTO, 3 TBD */
214 	void	*t_pspare2[4];		/* 1 TCP_SIGNATURE, 3 TBD */
215 	uint64_t _pad[6];		/* 6 TBD (1-2 CC/RTT?) */
216 };
217 
218 /*
219  * Flags and utility macros for the t_flags field.
220  */
221 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
222 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
223 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
224 #define	TF_NOOPT	0x000008	/* don't use tcp options */
225 #define	TF_SENTFIN	0x000010	/* have sent FIN */
226 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
227 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
228 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
229 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
230 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
231 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
232 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
233 #define	TF_NOPUSH	0x001000	/* don't push */
234 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
235 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
236 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
237 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
238 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
239 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
240 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
241 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
242 #define	TF_FORCEDATA	0x800000	/* force out a byte */
243 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
244 #define	TF_TOE		0x2000000	/* this connection is offloaded */
245 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
246 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
247 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
248 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
249 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
250 
251 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
252 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
253 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
254 
255 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
256 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
257 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
258 
259 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
260 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
261 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
262 
263 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
264 
265 /*
266  * Flags for the t_oobflags field.
267  */
268 #define	TCPOOB_HAVEDATA	0x01
269 #define	TCPOOB_HADDATA	0x02
270 
271 #ifdef TCP_SIGNATURE
272 /*
273  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
274  * for SADB verification and lookup.
275  */
276 #define	TCP_SIGLEN	16	/* length of computed digest in bytes */
277 #define	TCP_KEYLEN_MIN	1	/* minimum length of TCP-MD5 key */
278 #define	TCP_KEYLEN_MAX	80	/* maximum length of TCP-MD5 key */
279 /*
280  * Only a single SA per host may be specified at this time. An SPI is
281  * needed in order for the KEY_ALLOCSA() lookup to work.
282  */
283 #define	TCP_SIG_SPI	0x1000
284 #endif /* TCP_SIGNATURE */
285 
286 /*
287  * Structure to hold TCP options that are only used during segment
288  * processing (in tcp_input), but not held in the tcpcb.
289  * It's basically used to reduce the number of parameters
290  * to tcp_dooptions and tcp_addoptions.
291  * The binary order of the to_flags is relevant for packing of the
292  * options in tcp_addoptions.
293  */
294 struct tcpopt {
295 	u_int64_t	to_flags;	/* which options are present */
296 #define	TOF_MSS		0x0001		/* maximum segment size */
297 #define	TOF_SCALE	0x0002		/* window scaling */
298 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
299 #define	TOF_TS		0x0010		/* timestamp */
300 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
301 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
302 #define	TOF_MAXOPT	0x0100
303 	u_int32_t	to_tsval;	/* new timestamp */
304 	u_int32_t	to_tsecr;	/* reflected timestamp */
305 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
306 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
307 	u_int16_t	to_mss;		/* maximum segment size */
308 	u_int8_t	to_wscale;	/* window scaling */
309 	u_int8_t	to_nsacks;	/* number of SACK blocks */
310 	u_int32_t	to_spare;	/* UTO */
311 };
312 
313 /*
314  * Flags for tcp_dooptions.
315  */
316 #define	TO_SYN		0x01		/* parse SYN-only options */
317 
318 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
319 	u_long	rmx_mtu;	/* MTU for this path */
320 	u_long	rmx_ssthresh;	/* outbound gateway buffer limit */
321 	u_long	rmx_rtt;	/* estimated round trip time */
322 	u_long	rmx_rttvar;	/* estimated rtt variance */
323 	u_long	rmx_bandwidth;	/* estimated bandwidth */
324 	u_long	rmx_cwnd;	/* congestion window */
325 	u_long	rmx_sendpipe;   /* outbound delay-bandwidth product */
326 	u_long	rmx_recvpipe;   /* inbound delay-bandwidth product */
327 };
328 
329 /*
330  * Used by tcp_maxmtu() to communicate interface specific features
331  * and limits at the time of connection setup.
332  */
333 struct tcp_ifcap {
334 	int	ifcap;
335 	u_int	tsomax;
336 };
337 
338 #ifndef _NETINET_IN_PCB_H_
339 struct in_conninfo;
340 #endif /* _NETINET_IN_PCB_H_ */
341 
342 struct tcptw {
343 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
344 	tcp_seq		snd_nxt;
345 	tcp_seq		rcv_nxt;
346 	tcp_seq		iss;
347 	tcp_seq		irs;
348 	u_short		last_win;	/* cached window value */
349 	u_short		tw_so_options;	/* copy of so_options */
350 	struct ucred	*tw_cred;	/* user credentials */
351 	u_int32_t	t_recent;
352 	u_int32_t	ts_offset;	/* our timestamp offset */
353 	u_int		t_starttime;
354 	int		tw_time;
355 	TAILQ_ENTRY(tcptw) tw_2msl;
356 	void		*tw_pspare;	/* TCP_SIGNATURE */
357 	u_int		*tw_spare;	/* TCP_SIGNATURE */
358 };
359 
360 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
361 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
362 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
363 
364 /*
365  * The smoothed round-trip time and estimated variance
366  * are stored as fixed point numbers scaled by the values below.
367  * For convenience, these scales are also used in smoothing the average
368  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
369  * With these scales, srtt has 3 bits to the right of the binary point,
370  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
371  * binary point, and is smoothed with an ALPHA of 0.75.
372  */
373 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
374 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
375 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
376 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
377 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
378 
379 /*
380  * The initial retransmission should happen at rtt + 4 * rttvar.
381  * Because of the way we do the smoothing, srtt and rttvar
382  * will each average +1/2 tick of bias.  When we compute
383  * the retransmit timer, we want 1/2 tick of rounding and
384  * 1 extra tick because of +-1/2 tick uncertainty in the
385  * firing of the timer.  The bias will give us exactly the
386  * 1.5 tick we need.  But, because the bias is
387  * statistical, we have to test that we don't drop below
388  * the minimum feasible timer (which is 2 ticks).
389  * This version of the macro adapted from a paper by Lawrence
390  * Brakmo and Larry Peterson which outlines a problem caused
391  * by insufficient precision in the original implementation,
392  * which results in inappropriately large RTO values for very
393  * fast networks.
394  */
395 #define	TCP_REXMTVAL(tp) \
396 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
397 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
398 
399 /*
400  * TCP statistics.
401  * Many of these should be kept per connection,
402  * but that's inconvenient at the moment.
403  */
404 struct	tcpstat {
405 	uint64_t tcps_connattempt;	/* connections initiated */
406 	uint64_t tcps_accepts;		/* connections accepted */
407 	uint64_t tcps_connects;		/* connections established */
408 	uint64_t tcps_drops;		/* connections dropped */
409 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
410 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
411 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
412 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
413 	uint64_t tcps_rttupdated;	/* times we succeeded */
414 	uint64_t tcps_delack;		/* delayed acks sent */
415 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
416 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
417 	uint64_t tcps_persisttimeo;	/* persist timeouts */
418 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
419 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
420 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
421 
422 	uint64_t tcps_sndtotal;		/* total packets sent */
423 	uint64_t tcps_sndpack;		/* data packets sent */
424 	uint64_t tcps_sndbyte;		/* data bytes sent */
425 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
426 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
427 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
428 	uint64_t tcps_sndacks;		/* ack-only packets sent */
429 	uint64_t tcps_sndprobe;		/* window probes sent */
430 	uint64_t tcps_sndurg;		/* packets sent with URG only */
431 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
432 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
433 
434 	uint64_t tcps_rcvtotal;		/* total packets received */
435 	uint64_t tcps_rcvpack;		/* packets received in sequence */
436 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
437 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
438 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
439 	uint64_t tcps_rcvmemdrop;	/* packets dropped for lack of memory */
440 	uint64_t tcps_rcvshort;		/* packets received too short */
441 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
442 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
443 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
444 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
445 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
446 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
447 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
448 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
449 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
450 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
451 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
452 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
453 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
454 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
455 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
456 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
457 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
458 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
459 	uint64_t tcps_pcbcachemiss;
460 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
461 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
462 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
463 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
464 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
465 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
466 	uint64_t tcps_persistdrop;	/* timeout in persist state */
467 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
468 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
469 	uint64_t tcps_listendrop;	/* listen queue overflows */
470 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
471 
472 	uint64_t tcps_sc_added;		/* entry added to syncache */
473 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
474 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
475 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
476 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
477 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
478 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
479 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
480 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
481 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
482 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
483 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
484 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
485 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
486 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
487 
488 	uint64_t tcps_hc_added;		/* entry added to hostcache */
489 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
490 
491 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
492 
493 	/* SACK related stats */
494 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
495 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
496 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
497 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
498 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
499 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
500 
501 	/* ECN related stats */
502 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
503 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
504 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
505 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
506 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
507 
508 	/* TCP_SIGNATURE related stats */
509 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
510 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
511 	uint64_t tcps_sig_err_buildsig;	/* Mismatching signature received */
512 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
513 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
514 
515 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
516 };
517 
518 #ifdef _KERNEL
519 #include <sys/counter.h>
520 
521 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
522 /*
523  * In-kernel consumers can use these accessor macros directly to update
524  * stats.
525  */
526 #define	TCPSTAT_ADD(name, val)	\
527     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
528 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
529 
530 /*
531  * Kernel module consumers must use this accessor macro.
532  */
533 void	kmod_tcpstat_inc(int statnum);
534 #define	KMOD_TCPSTAT_INC(name)						\
535     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
536 
537 /*
538  * TCP specific helper hook point identifiers.
539  */
540 #define	HHOOK_TCP_EST_IN		0
541 #define	HHOOK_TCP_EST_OUT		1
542 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
543 
544 struct tcp_hhook_data {
545 	struct tcpcb	*tp;
546 	struct tcphdr	*th;
547 	struct tcpopt	*to;
548 	long		len;
549 	int		tso;
550 	tcp_seq		curack;
551 };
552 #endif
553 
554 /*
555  * TCB structure exported to user-land via sysctl(3).
556  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
557  * included.  Not all of our clients do.
558  */
559 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
560 struct xtcp_timer {
561 	int tt_rexmt;	/* retransmit timer */
562 	int tt_persist;	/* retransmit persistence */
563 	int tt_keep;	/* keepalive */
564 	int tt_2msl;	/* 2*msl TIME_WAIT timer */
565 	int tt_delack;	/* delayed ACK timer */
566 	int t_rcvtime;	/* Time since last packet received */
567 };
568 struct	xtcpcb {
569 	size_t	xt_len;
570 	struct	inpcb	xt_inp;
571 	struct	tcpcb	xt_tp;
572 	struct	xsocket	xt_socket;
573 	struct	xtcp_timer xt_timer;
574 	u_quad_t	xt_alignment_hack;
575 };
576 #endif
577 
578 /*
579  * Identifiers for TCP sysctl nodes
580  */
581 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
582 #define	TCPCTL_MSSDFLT		3	/* MSS default */
583 #define TCPCTL_STATS		4	/* statistics (read-only) */
584 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
585 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
586 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
587 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
588 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
589 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
590 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
591 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
592 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
593 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
594 #define	TCPCTL_DROP		15	/* drop tcp connection */
595 
596 #ifdef _KERNEL
597 #ifdef SYSCTL_DECL
598 SYSCTL_DECL(_net_inet_tcp);
599 SYSCTL_DECL(_net_inet_tcp_sack);
600 MALLOC_DECLARE(M_TCPLOG);
601 #endif
602 
603 VNET_DECLARE(struct inpcbhead, tcb);		/* queue of active tcpcb's */
604 VNET_DECLARE(struct inpcbinfo, tcbinfo);
605 extern	int tcp_log_in_vain;
606 VNET_DECLARE(int, tcp_mssdflt);	/* XXX */
607 VNET_DECLARE(int, tcp_minmss);
608 VNET_DECLARE(int, tcp_delack_enabled);
609 VNET_DECLARE(int, tcp_do_rfc3390);
610 VNET_DECLARE(int, tcp_do_initcwnd10);
611 VNET_DECLARE(int, tcp_sendspace);
612 VNET_DECLARE(int, tcp_recvspace);
613 VNET_DECLARE(int, path_mtu_discovery);
614 VNET_DECLARE(int, tcp_do_rfc3465);
615 VNET_DECLARE(int, tcp_abc_l_var);
616 #define	V_tcb			VNET(tcb)
617 #define	V_tcbinfo		VNET(tcbinfo)
618 #define	V_tcp_mssdflt		VNET(tcp_mssdflt)
619 #define	V_tcp_minmss		VNET(tcp_minmss)
620 #define	V_tcp_delack_enabled	VNET(tcp_delack_enabled)
621 #define	V_tcp_do_rfc3390	VNET(tcp_do_rfc3390)
622 #define	V_tcp_do_initcwnd10	VNET(tcp_do_initcwnd10)
623 #define	V_tcp_sendspace		VNET(tcp_sendspace)
624 #define	V_tcp_recvspace		VNET(tcp_recvspace)
625 #define	V_path_mtu_discovery	VNET(path_mtu_discovery)
626 #define	V_tcp_do_rfc3465	VNET(tcp_do_rfc3465)
627 #define	V_tcp_abc_l_var		VNET(tcp_abc_l_var)
628 
629 VNET_DECLARE(int, tcp_do_sack);			/* SACK enabled/disabled */
630 VNET_DECLARE(int, tcp_sc_rst_sock_fail);	/* RST on sock alloc failure */
631 #define	V_tcp_do_sack		VNET(tcp_do_sack)
632 #define	V_tcp_sc_rst_sock_fail	VNET(tcp_sc_rst_sock_fail)
633 
634 VNET_DECLARE(int, tcp_do_ecn);			/* TCP ECN enabled/disabled */
635 VNET_DECLARE(int, tcp_ecn_maxretries);
636 #define	V_tcp_do_ecn		VNET(tcp_do_ecn)
637 #define	V_tcp_ecn_maxretries	VNET(tcp_ecn_maxretries)
638 
639 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
640 #define	V_tcp_hhh		VNET(tcp_hhh)
641 
642 int	 tcp_addoptions(struct tcpopt *, u_char *);
643 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
644 struct tcpcb *
645 	 tcp_close(struct tcpcb *);
646 void	 tcp_discardcb(struct tcpcb *);
647 void	 tcp_twstart(struct tcpcb *);
648 #if 0
649 int	 tcp_twrecycleable(struct tcptw *tw);
650 #endif
651 void	 tcp_twclose(struct tcptw *_tw, int _reuse);
652 void	 tcp_ctlinput(int, struct sockaddr *, void *);
653 int	 tcp_ctloutput(struct socket *, struct sockopt *);
654 struct tcpcb *
655 	 tcp_drop(struct tcpcb *, int);
656 void	 tcp_drain(void);
657 void	 tcp_init(void);
658 #ifdef VIMAGE
659 void	 tcp_destroy(void);
660 #endif
661 void	 tcp_fini(void *);
662 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
663 	    const void *);
664 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
665 	    const void *);
666 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
667 void	 tcp_reass_init(void);
668 void	 tcp_reass_flush(struct tcpcb *);
669 #ifdef VIMAGE
670 void	 tcp_reass_destroy(void);
671 #endif
672 void	 tcp_input(struct mbuf *, int);
673 u_long	 tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
674 u_long	 tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
675 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
676 	    struct tcp_ifcap *);
677 void	 tcp_mss(struct tcpcb *, int);
678 int	 tcp_mssopt(struct in_conninfo *);
679 struct inpcb *
680 	 tcp_drop_syn_sent(struct inpcb *, int);
681 struct inpcb *
682 	 tcp_mtudisc(struct inpcb *, int);
683 struct tcpcb *
684 	 tcp_newtcpcb(struct inpcb *);
685 int	 tcp_output(struct tcpcb *);
686 void	 tcp_state_change(struct tcpcb *, int);
687 void	 tcp_respond(struct tcpcb *, void *,
688 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
689 void	 tcp_tw_init(void);
690 #ifdef VIMAGE
691 void	 tcp_tw_destroy(void);
692 #endif
693 void	 tcp_tw_zone_change(void);
694 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
695 	    struct mbuf *, int);
696 int	 tcp_twrespond(struct tcptw *, int);
697 void	 tcp_setpersist(struct tcpcb *);
698 #ifdef TCP_SIGNATURE
699 int	 tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int);
700 int	 tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *,
701 	    struct tcphdr *, u_int);
702 #endif
703 void	 tcp_slowtimo(void);
704 struct tcptemp *
705 	 tcpip_maketemplate(struct inpcb *);
706 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
707 void	 tcp_timer_activate(struct tcpcb *, int, u_int);
708 int	 tcp_timer_active(struct tcpcb *, int);
709 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
710 /*
711  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
712  */
713 void	 tcp_hc_init(void);
714 #ifdef VIMAGE
715 void	 tcp_hc_destroy(void);
716 #endif
717 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
718 u_long	 tcp_hc_getmtu(struct in_conninfo *);
719 void	 tcp_hc_updatemtu(struct in_conninfo *, u_long);
720 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
721 
722 extern	struct pr_usrreqs tcp_usrreqs;
723 tcp_seq tcp_new_isn(struct tcpcb *);
724 
725 void	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
726 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
727 void	 tcp_clean_sackreport(struct tcpcb *tp);
728 void	 tcp_sack_adjust(struct tcpcb *tp);
729 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
730 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
731 void	 tcp_free_sackholes(struct tcpcb *tp);
732 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
733 u_long	 tcp_seq_subtract(u_long, u_long );
734 
735 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
736 
737 #endif /* _KERNEL */
738 
739 #endif /* _NETINET_TCP_VAR_H_ */
740