xref: /freebsd/sys/netinet/tcp_var.h (revision 15c433351f54e7cd5bec8d36c8e89e6a7fa55b26)
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 #include <netinet/tcp_fsm.h>
38 
39 #ifdef _KERNEL
40 #include <net/vnet.h>
41 #include <sys/mbuf.h>
42 
43 /*
44  * Kernel variables for tcp.
45  */
46 VNET_DECLARE(int, tcp_do_rfc1323);
47 #define	V_tcp_do_rfc1323	VNET(tcp_do_rfc1323)
48 
49 #endif /* _KERNEL */
50 
51 /* TCP segment queue entry */
52 struct tseg_qent {
53 	LIST_ENTRY(tseg_qent) tqe_q;
54 	int	tqe_len;		/* TCP segment data length */
55 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
56 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
57 };
58 LIST_HEAD(tsegqe_head, tseg_qent);
59 
60 struct sackblk {
61 	tcp_seq start;		/* start seq no. of sack block */
62 	tcp_seq end;		/* end seq no. */
63 };
64 
65 struct sackhole {
66 	tcp_seq start;		/* start seq no. of hole */
67 	tcp_seq end;		/* end seq no. */
68 	tcp_seq rxmit;		/* next seq. no in hole to be retransmitted */
69 	TAILQ_ENTRY(sackhole) scblink;	/* scoreboard linkage */
70 };
71 
72 struct sackhint {
73 	struct sackhole	*nexthole;
74 	int		sack_bytes_rexmit;
75 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
76 
77 	int		ispare;		/* explicit pad for 64bit alignment */
78 	int             sacked_bytes;	/*
79 					 * Total sacked bytes reported by the
80 					 * receiver via sack option
81 					 */
82 	uint32_t	_pad1[1];	/* TBD */
83 	uint64_t	_pad[1];	/* TBD */
84 };
85 
86 struct tcptemp {
87 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
88 	struct	tcphdr tt_t;
89 };
90 
91 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
92 
93 /*
94  * TODO: We yet need to brave plowing in
95  * to tcp_input() and the pru_usrreq() block.
96  * Right now these go to the old standards which
97  * are somewhat ok, but in the long term may
98  * need to be changed. If we do tackle tcp_input()
99  * then we need to get rid of the tcp_do_segment()
100  * function below.
101  */
102 /* Flags for tcp functions */
103 #define TCP_FUNC_BEING_REMOVED 0x01   	/* Can no longer be referenced */
104 struct tcpcb;
105 struct inpcb;
106 struct sockopt;
107 struct socket;
108 
109 struct tcp_function_block {
110 	char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
111 	int	(*tfb_tcp_output)(struct tcpcb *);
112 	void	(*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
113 			    struct socket *, struct tcpcb *,
114 			    int, int, uint8_t,
115 			    int);
116 	int     (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt,
117 			    struct inpcb *inp, struct tcpcb *tp);
118 	/* Optional memory allocation/free routine */
119 	void	(*tfb_tcp_fb_init)(struct tcpcb *);
120 	void	(*tfb_tcp_fb_fini)(struct tcpcb *);
121 	/* Optional timers, must define all if you define one */
122 	int	(*tfb_tcp_timer_stop_all)(struct tcpcb *);
123 	int	(*tfb_tcp_timers_left)(struct tcpcb *);
124 	void	(*tfb_tcp_timer_activate)(struct tcpcb *,
125 			    uint32_t, u_int);
126 	int	(*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
127 	void	(*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
128 	volatile uint32_t tfb_refcnt;
129 	uint32_t  tfb_flags;
130 };
131 
132 struct tcp_function {
133 	TAILQ_ENTRY(tcp_function) tf_next;
134 	struct tcp_function_block *tf_fb;
135 };
136 
137 TAILQ_HEAD(tcp_funchead, tcp_function);
138 
139 /*
140  * Tcp control block, one per tcp; fields:
141  * Organized for 16 byte cacheline efficiency.
142  */
143 struct tcpcb {
144 	struct	tsegqe_head t_segq;	/* segment reassembly queue */
145 	void	*t_pspare[2];		/* new reassembly queue */
146 	int	t_segqlen;		/* segment reassembly queue length */
147 	int	t_dupacks;		/* consecutive dup acks recd */
148 
149 	struct tcp_timer *t_timers;	/* All the TCP timers in one struct */
150 
151 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
152 	int	t_state;		/* state of this connection */
153 	u_int	t_flags;
154 
155 	struct	vnet *t_vnet;		/* back pointer to parent vnet */
156 
157 	tcp_seq	snd_una;		/* sent but unacknowledged */
158 	tcp_seq	snd_max;		/* highest sequence number sent;
159 					 * used to recognize retransmits
160 					 */
161 	tcp_seq	snd_nxt;		/* send next */
162 	tcp_seq	snd_up;			/* send urgent pointer */
163 
164 	tcp_seq	snd_wl1;		/* window update seg seq number */
165 	tcp_seq	snd_wl2;		/* window update seg ack number */
166 	tcp_seq	iss;			/* initial send sequence number */
167 	tcp_seq	irs;			/* initial receive sequence number */
168 
169 	tcp_seq	rcv_nxt;		/* receive next */
170 	tcp_seq	rcv_adv;		/* advertised window */
171 	u_long	rcv_wnd;		/* receive window */
172 	tcp_seq	rcv_up;			/* receive urgent pointer */
173 
174 	u_long	snd_wnd;		/* send window */
175 	u_long	snd_cwnd;		/* congestion-controlled window */
176 	u_long	snd_spare1;		/* unused */
177 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
178 					 * for slow start exponential to
179 					 * linear switch
180 					 */
181 	u_long	snd_spare2;		/* unused */
182 	tcp_seq	snd_recover;		/* for use in NewReno Fast Recovery */
183 
184 	u_int	t_rcvtime;		/* inactivity time */
185 	u_int	t_starttime;		/* time connection was established */
186 	u_int	t_rtttime;		/* RTT measurement start time */
187 	tcp_seq	t_rtseq;		/* sequence number being timed */
188 
189 	u_int	t_bw_spare1;		/* unused */
190 	tcp_seq	t_bw_spare2;		/* unused */
191 
192 	int	t_rxtcur;		/* current retransmit value (ticks) */
193 	u_int	t_maxseg;		/* maximum segment size */
194 	u_int	t_pmtud_saved_maxseg;	/* pre-blackhole MSS */
195 	int	t_srtt;			/* smoothed round-trip time */
196 	int	t_rttvar;		/* variance in round-trip time */
197 
198 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
199 	u_int	t_rttmin;		/* minimum rtt allowed */
200 	u_int	t_rttbest;		/* best rtt we've seen */
201 	u_long	t_rttupdated;		/* number of times rtt sampled */
202 	u_long	max_sndwnd;		/* largest window peer has offered */
203 
204 	int	t_softerror;		/* possible error not yet reported */
205 /* out-of-band data */
206 	char	t_oobflags;		/* have some */
207 	char	t_iobc;			/* input character */
208 /* RFC 1323 variables */
209 	u_char	snd_scale;		/* window scaling for send window */
210 	u_char	rcv_scale;		/* window scaling for recv window */
211 	u_char	request_r_scale;	/* pending window scaling */
212 	u_int32_t  ts_recent;		/* timestamp echo data */
213 	u_int	ts_recent_age;		/* when last updated */
214 	u_int32_t  ts_offset;		/* our timestamp offset */
215 
216 	tcp_seq	last_ack_sent;
217 /* experimental */
218 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
219 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
220 	tcp_seq	snd_recover_prev;	/* snd_recover prior to retransmit */
221 	int	t_sndzerowin;		/* zero-window updates sent */
222 	u_int	t_badrxtwin;		/* window for retransmit recovery */
223 	u_char	snd_limited;		/* segments limited transmitted */
224 /* SACK related state */
225 	int	snd_numholes;		/* number of holes seen by sender */
226 	TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
227 					/* SACK scoreboard (sorted) */
228 	tcp_seq	snd_fack;		/* last seq number(+1) sack'd by rcv'r*/
229 	int	rcv_numsacks;		/* # distinct sack blks present */
230 	struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
231 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
232 					   episode starts at this seq number */
233 	struct sackhint	sackhint;	/* SACK scoreboard hint */
234 	int	t_rttlow;		/* smallest observerved RTT */
235 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
236 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
237 	struct toedev	*tod;		/* toedev handling this connection */
238 	int	t_sndrexmitpack;	/* retransmit packets sent */
239 	int	t_rcvoopack;		/* out-of-order packets received */
240 	void	*t_toe;			/* TOE pcb pointer */
241 	int	t_bytes_acked;		/* # bytes acked during current RTT */
242 	struct cc_algo	*cc_algo;	/* congestion control algorithm */
243 	struct cc_var	*ccv;		/* congestion control specific vars */
244 	struct osd	*osd;		/* storage for Khelp module data */
245 
246 	u_int	t_keepinit;		/* time to establish connection */
247 	u_int	t_keepidle;		/* time before keepalive probes begin */
248 	u_int	t_keepintvl;		/* interval between keepalives */
249 	u_int	t_keepcnt;		/* number of keepalives before close */
250 
251 	u_int	t_tsomax;		/* TSO total burst length limit in bytes */
252 	u_int	t_tsomaxsegcount;	/* TSO maximum segment count */
253 	u_int	t_tsomaxsegsize;	/* TSO maximum segment size in bytes */
254 	u_int	t_flags2;		/* More tcpcb flags storage */
255 #if defined(_KERNEL) && defined(TCP_RFC7413)
256 	uint32_t t_ispare[6];		/* 5 UTO, 1 TBD */
257 	uint64_t t_tfo_cookie;		/* TCP Fast Open cookie */
258 #else
259 	uint32_t t_ispare[8];		/* 5 UTO, 3 TBD */
260 #endif
261 	struct tcp_function_block *t_fb;/* TCP function call block */
262 	void	*t_fb_ptr;		/* Pointer to t_fb specific data */
263 #if defined(_KERNEL) && defined(TCP_RFC7413)
264 	unsigned int *t_tfo_pending;	/* TCP Fast Open pending counter */
265 	void	*t_pspare2[1];		/* 1 TCP_SIGNATURE */
266 #else
267 	void	*t_pspare2[2];		/* 1 TCP_SIGNATURE, 1 TBD */
268 #endif
269 #if defined(_KERNEL) && defined(TCPPCAP)
270 	struct mbufq t_inpkts;		/* List of saved input packets. */
271 	struct mbufq t_outpkts;		/* List of saved output packets. */
272 #ifdef _LP64
273 	uint64_t _pad[0];		/* all used! */
274 #else
275 	uint64_t _pad[2];		/* 2 are available */
276 #endif /* _LP64 */
277 #else
278 	uint64_t _pad[6];
279 #endif /* defined(_KERNEL) && defined(TCPPCAP) */
280 };
281 
282 /*
283  * Flags and utility macros for the t_flags field.
284  */
285 #define	TF_ACKNOW	0x000001	/* ack peer immediately */
286 #define	TF_DELACK	0x000002	/* ack, but try to delay it */
287 #define	TF_NODELAY	0x000004	/* don't delay packets to coalesce */
288 #define	TF_NOOPT	0x000008	/* don't use tcp options */
289 #define	TF_SENTFIN	0x000010	/* have sent FIN */
290 #define	TF_REQ_SCALE	0x000020	/* have/will request window scaling */
291 #define	TF_RCVD_SCALE	0x000040	/* other side has requested scaling */
292 #define	TF_REQ_TSTMP	0x000080	/* have/will request timestamps */
293 #define	TF_RCVD_TSTMP	0x000100	/* a timestamp was received in SYN */
294 #define	TF_SACK_PERMIT	0x000200	/* other side said I could SACK */
295 #define	TF_NEEDSYN	0x000400	/* send SYN (implicit state) */
296 #define	TF_NEEDFIN	0x000800	/* send FIN (implicit state) */
297 #define	TF_NOPUSH	0x001000	/* don't push */
298 #define	TF_PREVVALID	0x002000	/* saved values for bad rxmit valid */
299 #define	TF_MORETOCOME	0x010000	/* More data to be appended to sock */
300 #define	TF_LQ_OVERFLOW	0x020000	/* listen queue overflow */
301 #define	TF_LASTIDLE	0x040000	/* connection was previously idle */
302 #define	TF_RXWIN0SENT	0x080000	/* sent a receiver win 0 in response */
303 #define	TF_FASTRECOVERY	0x100000	/* in NewReno Fast Recovery */
304 #define	TF_WASFRECOVERY	0x200000	/* was in NewReno Fast Recovery */
305 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
306 #define	TF_FORCEDATA	0x800000	/* force out a byte */
307 #define	TF_TSO		0x1000000	/* TSO enabled on this connection */
308 #define	TF_TOE		0x2000000	/* this connection is offloaded */
309 #define	TF_ECN_PERMIT	0x4000000	/* connection ECN-ready */
310 #define	TF_ECN_SND_CWR	0x8000000	/* ECN CWR in queue */
311 #define	TF_ECN_SND_ECE	0x10000000	/* ECN ECE in queue */
312 #define	TF_CONGRECOVERY	0x20000000	/* congestion recovery mode */
313 #define	TF_WASCRECOVERY	0x40000000	/* was in congestion recovery */
314 #define	TF_FASTOPEN	0x80000000	/* TCP Fast Open indication */
315 
316 #define	IN_FASTRECOVERY(t_flags)	(t_flags & TF_FASTRECOVERY)
317 #define	ENTER_FASTRECOVERY(t_flags)	t_flags |= TF_FASTRECOVERY
318 #define	EXIT_FASTRECOVERY(t_flags)	t_flags &= ~TF_FASTRECOVERY
319 
320 #define	IN_CONGRECOVERY(t_flags)	(t_flags & TF_CONGRECOVERY)
321 #define	ENTER_CONGRECOVERY(t_flags)	t_flags |= TF_CONGRECOVERY
322 #define	EXIT_CONGRECOVERY(t_flags)	t_flags &= ~TF_CONGRECOVERY
323 
324 #define	IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
325 #define	ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
326 #define	EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
327 
328 #define	BYTES_THIS_ACK(tp, th)	(th->th_ack - tp->snd_una)
329 
330 /*
331  * Flags for the t_oobflags field.
332  */
333 #define	TCPOOB_HAVEDATA	0x01
334 #define	TCPOOB_HADDATA	0x02
335 
336 #ifdef TCP_SIGNATURE
337 /*
338  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
339  * for SADB verification and lookup.
340  */
341 #define	TCP_SIGLEN	16	/* length of computed digest in bytes */
342 #define	TCP_KEYLEN_MIN	1	/* minimum length of TCP-MD5 key */
343 #define	TCP_KEYLEN_MAX	80	/* maximum length of TCP-MD5 key */
344 /*
345  * Only a single SA per host may be specified at this time. An SPI is
346  * needed in order for the KEY_ALLOCSA() lookup to work.
347  */
348 #define	TCP_SIG_SPI	0x1000
349 #endif /* TCP_SIGNATURE */
350 
351 /*
352  * Flags for PLPMTU handling, t_flags2
353  */
354 #define	TF2_PLPMTU_BLACKHOLE	0x00000001 /* Possible PLPMTUD Black Hole. */
355 #define	TF2_PLPMTU_PMTUD	0x00000002 /* Allowed to attempt PLPMTUD. */
356 #define	TF2_PLPMTU_MAXSEGSNT	0x00000004 /* Last seg sent was full seg. */
357 
358 /*
359  * Structure to hold TCP options that are only used during segment
360  * processing (in tcp_input), but not held in the tcpcb.
361  * It's basically used to reduce the number of parameters
362  * to tcp_dooptions and tcp_addoptions.
363  * The binary order of the to_flags is relevant for packing of the
364  * options in tcp_addoptions.
365  */
366 struct tcpopt {
367 	u_int64_t	to_flags;	/* which options are present */
368 #define	TOF_MSS		0x0001		/* maximum segment size */
369 #define	TOF_SCALE	0x0002		/* window scaling */
370 #define	TOF_SACKPERM	0x0004		/* SACK permitted */
371 #define	TOF_TS		0x0010		/* timestamp */
372 #define	TOF_SIGNATURE	0x0040		/* TCP-MD5 signature option (RFC2385) */
373 #define	TOF_SACK	0x0080		/* Peer sent SACK option */
374 #define	TOF_FASTOPEN	0x0100		/* TCP Fast Open (TFO) cookie */
375 #define	TOF_MAXOPT	0x0200
376 	u_int32_t	to_tsval;	/* new timestamp */
377 	u_int32_t	to_tsecr;	/* reflected timestamp */
378 	u_char		*to_sacks;	/* pointer to the first SACK blocks */
379 	u_char		*to_signature;	/* pointer to the TCP-MD5 signature */
380 	u_char		*to_tfo_cookie; /* pointer to the TFO cookie */
381 	u_int16_t	to_mss;		/* maximum segment size */
382 	u_int8_t	to_wscale;	/* window scaling */
383 	u_int8_t	to_nsacks;	/* number of SACK blocks */
384 	u_int8_t	to_tfo_len;	/* TFO cookie length */
385 	u_int32_t	to_spare;	/* UTO */
386 };
387 
388 /*
389  * Flags for tcp_dooptions.
390  */
391 #define	TO_SYN		0x01		/* parse SYN-only options */
392 
393 struct hc_metrics_lite {	/* must stay in sync with hc_metrics */
394 	u_long	rmx_mtu;	/* MTU for this path */
395 	u_long	rmx_ssthresh;	/* outbound gateway buffer limit */
396 	u_long	rmx_rtt;	/* estimated round trip time */
397 	u_long	rmx_rttvar;	/* estimated rtt variance */
398 	u_long	rmx_cwnd;	/* congestion window */
399 	u_long	rmx_sendpipe;   /* outbound delay-bandwidth product */
400 	u_long	rmx_recvpipe;   /* inbound delay-bandwidth product */
401 };
402 
403 /*
404  * Used by tcp_maxmtu() to communicate interface specific features
405  * and limits at the time of connection setup.
406  */
407 struct tcp_ifcap {
408 	int	ifcap;
409 	u_int	tsomax;
410 	u_int	tsomaxsegcount;
411 	u_int	tsomaxsegsize;
412 };
413 
414 #ifndef _NETINET_IN_PCB_H_
415 struct in_conninfo;
416 #endif /* _NETINET_IN_PCB_H_ */
417 
418 struct tcptw {
419 	struct inpcb	*tw_inpcb;	/* XXX back pointer to internet pcb */
420 	tcp_seq		snd_nxt;
421 	tcp_seq		rcv_nxt;
422 	tcp_seq		iss;
423 	tcp_seq		irs;
424 	u_short		last_win;	/* cached window value */
425 	u_short		tw_so_options;	/* copy of so_options */
426 	struct ucred	*tw_cred;	/* user credentials */
427 	u_int32_t	t_recent;
428 	u_int32_t	ts_offset;	/* our timestamp offset */
429 	u_int		t_starttime;
430 	int		tw_time;
431 	TAILQ_ENTRY(tcptw) tw_2msl;
432 	void		*tw_pspare;	/* TCP_SIGNATURE */
433 	u_int		*tw_spare;	/* TCP_SIGNATURE */
434 };
435 
436 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
437 #define	intotw(ip)	((struct tcptw *)(ip)->inp_ppcb)
438 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
439 
440 /*
441  * The smoothed round-trip time and estimated variance
442  * are stored as fixed point numbers scaled by the values below.
443  * For convenience, these scales are also used in smoothing the average
444  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
445  * With these scales, srtt has 3 bits to the right of the binary point,
446  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
447  * binary point, and is smoothed with an ALPHA of 0.75.
448  */
449 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
450 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
451 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
452 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
453 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
454 
455 /*
456  * The initial retransmission should happen at rtt + 4 * rttvar.
457  * Because of the way we do the smoothing, srtt and rttvar
458  * will each average +1/2 tick of bias.  When we compute
459  * the retransmit timer, we want 1/2 tick of rounding and
460  * 1 extra tick because of +-1/2 tick uncertainty in the
461  * firing of the timer.  The bias will give us exactly the
462  * 1.5 tick we need.  But, because the bias is
463  * statistical, we have to test that we don't drop below
464  * the minimum feasible timer (which is 2 ticks).
465  * This version of the macro adapted from a paper by Lawrence
466  * Brakmo and Larry Peterson which outlines a problem caused
467  * by insufficient precision in the original implementation,
468  * which results in inappropriately large RTO values for very
469  * fast networks.
470  */
471 #define	TCP_REXMTVAL(tp) \
472 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
473 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
474 
475 /*
476  * TCP statistics.
477  * Many of these should be kept per connection,
478  * but that's inconvenient at the moment.
479  */
480 struct	tcpstat {
481 	uint64_t tcps_connattempt;	/* connections initiated */
482 	uint64_t tcps_accepts;		/* connections accepted */
483 	uint64_t tcps_connects;		/* connections established */
484 	uint64_t tcps_drops;		/* connections dropped */
485 	uint64_t tcps_conndrops;	/* embryonic connections dropped */
486 	uint64_t tcps_minmssdrops;	/* average minmss too low drops */
487 	uint64_t tcps_closed;		/* conn. closed (includes drops) */
488 	uint64_t tcps_segstimed;	/* segs where we tried to get rtt */
489 	uint64_t tcps_rttupdated;	/* times we succeeded */
490 	uint64_t tcps_delack;		/* delayed acks sent */
491 	uint64_t tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
492 	uint64_t tcps_rexmttimeo;	/* retransmit timeouts */
493 	uint64_t tcps_persisttimeo;	/* persist timeouts */
494 	uint64_t tcps_keeptimeo;	/* keepalive timeouts */
495 	uint64_t tcps_keepprobe;	/* keepalive probes sent */
496 	uint64_t tcps_keepdrops;	/* connections dropped in keepalive */
497 
498 	uint64_t tcps_sndtotal;		/* total packets sent */
499 	uint64_t tcps_sndpack;		/* data packets sent */
500 	uint64_t tcps_sndbyte;		/* data bytes sent */
501 	uint64_t tcps_sndrexmitpack;	/* data packets retransmitted */
502 	uint64_t tcps_sndrexmitbyte;	/* data bytes retransmitted */
503 	uint64_t tcps_sndrexmitbad;	/* unnecessary packet retransmissions */
504 	uint64_t tcps_sndacks;		/* ack-only packets sent */
505 	uint64_t tcps_sndprobe;		/* window probes sent */
506 	uint64_t tcps_sndurg;		/* packets sent with URG only */
507 	uint64_t tcps_sndwinup;		/* window update-only packets sent */
508 	uint64_t tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
509 
510 	uint64_t tcps_rcvtotal;		/* total packets received */
511 	uint64_t tcps_rcvpack;		/* packets received in sequence */
512 	uint64_t tcps_rcvbyte;		/* bytes received in sequence */
513 	uint64_t tcps_rcvbadsum;	/* packets received with ccksum errs */
514 	uint64_t tcps_rcvbadoff;	/* packets received with bad offset */
515 	uint64_t tcps_rcvreassfull;	/* packets dropped for no reass space */
516 	uint64_t tcps_rcvshort;		/* packets received too short */
517 	uint64_t tcps_rcvduppack;	/* duplicate-only packets received */
518 	uint64_t tcps_rcvdupbyte;	/* duplicate-only bytes received */
519 	uint64_t tcps_rcvpartduppack;	/* packets with some duplicate data */
520 	uint64_t tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
521 	uint64_t tcps_rcvoopack;	/* out-of-order packets received */
522 	uint64_t tcps_rcvoobyte;	/* out-of-order bytes received */
523 	uint64_t tcps_rcvpackafterwin;	/* packets with data after window */
524 	uint64_t tcps_rcvbyteafterwin;	/* bytes rcvd after window */
525 	uint64_t tcps_rcvafterclose;	/* packets rcvd after "close" */
526 	uint64_t tcps_rcvwinprobe;	/* rcvd window probe packets */
527 	uint64_t tcps_rcvdupack;	/* rcvd duplicate acks */
528 	uint64_t tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
529 	uint64_t tcps_rcvackpack;	/* rcvd ack packets */
530 	uint64_t tcps_rcvackbyte;	/* bytes acked by rcvd acks */
531 	uint64_t tcps_rcvwinupd;	/* rcvd window update packets */
532 	uint64_t tcps_pawsdrop;		/* segments dropped due to PAWS */
533 	uint64_t tcps_predack;		/* times hdr predict ok for acks */
534 	uint64_t tcps_preddat;		/* times hdr predict ok for data pkts */
535 	uint64_t tcps_pcbcachemiss;
536 	uint64_t tcps_cachedrtt;	/* times cached RTT in route updated */
537 	uint64_t tcps_cachedrttvar;	/* times cached rttvar updated */
538 	uint64_t tcps_cachedssthresh;	/* times cached ssthresh updated */
539 	uint64_t tcps_usedrtt;		/* times RTT initialized from route */
540 	uint64_t tcps_usedrttvar;	/* times RTTVAR initialized from rt */
541 	uint64_t tcps_usedssthresh;	/* times ssthresh initialized from rt*/
542 	uint64_t tcps_persistdrop;	/* timeout in persist state */
543 	uint64_t tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
544 	uint64_t tcps_mturesent;	/* resends due to MTU discovery */
545 	uint64_t tcps_listendrop;	/* listen queue overflows */
546 	uint64_t tcps_badrst;		/* ignored RSTs in the window */
547 
548 	uint64_t tcps_sc_added;		/* entry added to syncache */
549 	uint64_t tcps_sc_retransmitted;	/* syncache entry was retransmitted */
550 	uint64_t tcps_sc_dupsyn;	/* duplicate SYN packet */
551 	uint64_t tcps_sc_dropped;	/* could not reply to packet */
552 	uint64_t tcps_sc_completed;	/* successful extraction of entry */
553 	uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
554 	uint64_t tcps_sc_cacheoverflow;	/* syncache cache limit hit */
555 	uint64_t tcps_sc_reset;		/* RST removed entry from syncache */
556 	uint64_t tcps_sc_stale;		/* timed out or listen socket gone */
557 	uint64_t tcps_sc_aborted;	/* syncache entry aborted */
558 	uint64_t tcps_sc_badack;	/* removed due to bad ACK */
559 	uint64_t tcps_sc_unreach;	/* ICMP unreachable received */
560 	uint64_t tcps_sc_zonefail;	/* zalloc() failed */
561 	uint64_t tcps_sc_sendcookie;	/* SYN cookie sent */
562 	uint64_t tcps_sc_recvcookie;	/* SYN cookie received */
563 
564 	uint64_t tcps_hc_added;		/* entry added to hostcache */
565 	uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
566 
567 	uint64_t tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
568 
569 	/* SACK related stats */
570 	uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
571 	uint64_t tcps_sack_rexmits;	    /* SACK rexmit segments   */
572 	uint64_t tcps_sack_rexmit_bytes;    /* SACK rexmit bytes      */
573 	uint64_t tcps_sack_rcv_blocks;	    /* SACK blocks (options) received */
574 	uint64_t tcps_sack_send_blocks;	    /* SACK blocks (options) sent     */
575 	uint64_t tcps_sack_sboverflow;	    /* times scoreboard overflowed */
576 
577 	/* ECN related stats */
578 	uint64_t tcps_ecn_ce;		/* ECN Congestion Experienced */
579 	uint64_t tcps_ecn_ect0;		/* ECN Capable Transport */
580 	uint64_t tcps_ecn_ect1;		/* ECN Capable Transport */
581 	uint64_t tcps_ecn_shs;		/* ECN successful handshakes */
582 	uint64_t tcps_ecn_rcwnd;	/* # times ECN reduced the cwnd */
583 
584 	/* TCP_SIGNATURE related stats */
585 	uint64_t tcps_sig_rcvgoodsig;	/* Total matching signature received */
586 	uint64_t tcps_sig_rcvbadsig;	/* Total bad signature received */
587 	uint64_t tcps_sig_err_buildsig;	/* Mismatching signature received */
588 	uint64_t tcps_sig_err_sigopt;	/* No signature expected by socket */
589 	uint64_t tcps_sig_err_nosigopt;	/* No signature provided by segment */
590 
591 	/* Running connection count. */
592 	uint64_t tcps_states[TCP_NSTATES];
593 
594 	uint64_t _pad[12];		/* 6 UTO, 6 TBD */
595 };
596 
597 #define	tcps_rcvmemdrop	tcps_rcvreassfull	/* compat */
598 
599 #ifdef _KERNEL
600 #define	TI_UNLOCKED	1
601 #define	TI_RLOCKED	2
602 #include <sys/counter.h>
603 
604 VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat);	/* tcp statistics */
605 /*
606  * In-kernel consumers can use these accessor macros directly to update
607  * stats.
608  */
609 #define	TCPSTAT_ADD(name, val)	\
610     VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val))
611 #define	TCPSTAT_INC(name)	TCPSTAT_ADD(name, 1)
612 #define	TCPSTAT_DEC(name)	TCPSTAT_ADD(name, -1)
613 #define	TCPSTAT_FETCH(name)	VNET_PCPUSTAT_FETCH(struct tcpstat, tcpstat, \
614 				    name)
615 
616 /*
617  * Kernel module consumers must use this accessor macro.
618  */
619 void	kmod_tcpstat_inc(int statnum);
620 #define	KMOD_TCPSTAT_INC(name)						\
621     kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t))
622 
623 /*
624  * TCP specific helper hook point identifiers.
625  */
626 #define	HHOOK_TCP_EST_IN		0
627 #define	HHOOK_TCP_EST_OUT		1
628 #define	HHOOK_TCP_LAST			HHOOK_TCP_EST_OUT
629 
630 struct tcp_hhook_data {
631 	struct tcpcb	*tp;
632 	struct tcphdr	*th;
633 	struct tcpopt	*to;
634 	long		len;
635 	int		tso;
636 	tcp_seq		curack;
637 };
638 #endif
639 
640 /*
641  * TCB structure exported to user-land via sysctl(3).
642  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
643  * included.  Not all of our clients do.
644  */
645 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
646 struct xtcp_timer {
647 	int tt_rexmt;	/* retransmit timer */
648 	int tt_persist;	/* retransmit persistence */
649 	int tt_keep;	/* keepalive */
650 	int tt_2msl;	/* 2*msl TIME_WAIT timer */
651 	int tt_delack;	/* delayed ACK timer */
652 	int t_rcvtime;	/* Time since last packet received */
653 };
654 struct	xtcpcb {
655 	size_t	xt_len;
656 	struct	inpcb	xt_inp;
657 	struct	tcpcb	xt_tp;
658 	struct	xsocket	xt_socket;
659 	struct	xtcp_timer xt_timer;
660 	u_quad_t	xt_alignment_hack;
661 };
662 #endif
663 
664 /*
665  * Identifiers for TCP sysctl nodes
666  */
667 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
668 #define	TCPCTL_MSSDFLT		3	/* MSS default */
669 #define TCPCTL_STATS		4	/* statistics (read-only) */
670 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
671 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
672 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
673 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
674 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
675 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
676 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
677 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
678 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
679 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
680 #define	TCPCTL_DROP		15	/* drop tcp connection */
681 
682 #ifdef _KERNEL
683 #ifdef SYSCTL_DECL
684 SYSCTL_DECL(_net_inet_tcp);
685 SYSCTL_DECL(_net_inet_tcp_sack);
686 MALLOC_DECLARE(M_TCPLOG);
687 #endif
688 
689 VNET_DECLARE(struct inpcbhead, tcb);		/* queue of active tcpcb's */
690 VNET_DECLARE(struct inpcbinfo, tcbinfo);
691 extern	int tcp_log_in_vain;
692 VNET_DECLARE(int, tcp_mssdflt);	/* XXX */
693 VNET_DECLARE(int, tcp_minmss);
694 VNET_DECLARE(int, tcp_delack_enabled);
695 VNET_DECLARE(int, tcp_do_rfc3390);
696 VNET_DECLARE(int, tcp_initcwnd_segments);
697 VNET_DECLARE(int, tcp_sendspace);
698 VNET_DECLARE(int, tcp_recvspace);
699 VNET_DECLARE(int, path_mtu_discovery);
700 VNET_DECLARE(int, tcp_do_rfc3465);
701 VNET_DECLARE(int, tcp_abc_l_var);
702 #define	V_tcb			VNET(tcb)
703 #define	V_tcbinfo		VNET(tcbinfo)
704 #define	V_tcp_mssdflt		VNET(tcp_mssdflt)
705 #define	V_tcp_minmss		VNET(tcp_minmss)
706 #define	V_tcp_delack_enabled	VNET(tcp_delack_enabled)
707 #define	V_tcp_do_rfc3390	VNET(tcp_do_rfc3390)
708 #define	V_tcp_initcwnd_segments	VNET(tcp_initcwnd_segments)
709 #define	V_tcp_sendspace		VNET(tcp_sendspace)
710 #define	V_tcp_recvspace		VNET(tcp_recvspace)
711 #define	V_path_mtu_discovery	VNET(path_mtu_discovery)
712 #define	V_tcp_do_rfc3465	VNET(tcp_do_rfc3465)
713 #define	V_tcp_abc_l_var		VNET(tcp_abc_l_var)
714 
715 VNET_DECLARE(int, tcp_do_sack);			/* SACK enabled/disabled */
716 VNET_DECLARE(int, tcp_sc_rst_sock_fail);	/* RST on sock alloc failure */
717 #define	V_tcp_do_sack		VNET(tcp_do_sack)
718 #define	V_tcp_sc_rst_sock_fail	VNET(tcp_sc_rst_sock_fail)
719 
720 VNET_DECLARE(int, tcp_do_ecn);			/* TCP ECN enabled/disabled */
721 VNET_DECLARE(int, tcp_ecn_maxretries);
722 #define	V_tcp_do_ecn		VNET(tcp_do_ecn)
723 #define	V_tcp_ecn_maxretries	VNET(tcp_ecn_maxretries)
724 
725 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
726 #define	V_tcp_hhh		VNET(tcp_hhh)
727 
728 VNET_DECLARE(int, tcp_do_rfc6675_pipe);
729 #define V_tcp_do_rfc6675_pipe	VNET(tcp_do_rfc6675_pipe)
730 
731 int	 tcp_addoptions(struct tcpopt *, u_char *);
732 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
733 struct tcpcb *
734 	 tcp_close(struct tcpcb *);
735 void	 tcp_discardcb(struct tcpcb *);
736 void	 tcp_twstart(struct tcpcb *);
737 void	 tcp_twclose(struct tcptw *, int);
738 void	 tcp_ctlinput(int, struct sockaddr *, void *);
739 int	 tcp_ctloutput(struct socket *, struct sockopt *);
740 struct tcpcb *
741 	 tcp_drop(struct tcpcb *, int);
742 void	 tcp_drain(void);
743 void	 tcp_init(void);
744 #ifdef VIMAGE
745 void	 tcp_destroy(void);
746 #endif
747 void	 tcp_fini(void *);
748 char	*tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *,
749 	    const void *);
750 char	*tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *,
751 	    const void *);
752 int	 tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *);
753 void	 tcp_reass_global_init(void);
754 void	 tcp_reass_flush(struct tcpcb *);
755 void	 tcp_dooptions(struct tcpopt *, u_char *, int, int);
756 void	tcp_dropwithreset(struct mbuf *, struct tcphdr *,
757 		     struct tcpcb *, int, int);
758 void	tcp_pulloutofband(struct socket *,
759 		     struct tcphdr *, struct mbuf *, int);
760 void	tcp_xmit_timer(struct tcpcb *, int);
761 void	tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
762 void	cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
763 			    uint16_t type);
764 void 	cc_conn_init(struct tcpcb *tp);
765 void 	cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
766 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
767 void	hhook_run_tcp_est_in(struct tcpcb *tp,
768 			    struct tcphdr *th, struct tcpopt *to);
769 
770 int	 tcp_input(struct mbuf **, int *, int);
771 void	 tcp_do_segment(struct mbuf *, struct tcphdr *,
772 			struct socket *, struct tcpcb *, int, int, uint8_t,
773 			int);
774 
775 int register_tcp_functions(struct tcp_function_block *blk, int wait);
776 int deregister_tcp_functions(struct tcp_function_block *blk);
777 struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
778 struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk);
779 int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp);
780 
781 u_long	 tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
782 u_long	 tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
783 u_int	 tcp_maxseg(const struct tcpcb *);
784 void	 tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
785 	    struct tcp_ifcap *);
786 void	 tcp_mss(struct tcpcb *, int);
787 int	 tcp_mssopt(struct in_conninfo *);
788 struct inpcb *
789 	 tcp_drop_syn_sent(struct inpcb *, int);
790 struct tcpcb *
791 	 tcp_newtcpcb(struct inpcb *);
792 int	 tcp_output(struct tcpcb *);
793 void	 tcp_state_change(struct tcpcb *, int);
794 void	 tcp_respond(struct tcpcb *, void *,
795 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
796 void	 tcp_tw_init(void);
797 #ifdef VIMAGE
798 void	 tcp_tw_destroy(void);
799 #endif
800 void	 tcp_tw_zone_change(void);
801 int	 tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
802 	    struct mbuf *, int);
803 void	 tcp_setpersist(struct tcpcb *);
804 #ifdef TCP_SIGNATURE
805 struct secasvar;
806 struct secasvar *tcp_get_sav(struct mbuf *, u_int);
807 int	 tcp_signature_do_compute(struct mbuf *, int, int, u_char *,
808 	    struct secasvar *);
809 int	 tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int);
810 int	 tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *,
811 	    struct tcphdr *, u_int);
812 int	tcp_signature_check(struct mbuf *m, int off0, int tlen, int optlen,
813 	    struct tcpopt *to, struct tcphdr *th, u_int tcpbflag);
814 #endif
815 void	 tcp_slowtimo(void);
816 struct tcptemp *
817 	 tcpip_maketemplate(struct inpcb *);
818 void	 tcpip_fillheaders(struct inpcb *, void *, void *);
819 void	 tcp_timer_activate(struct tcpcb *, uint32_t, u_int);
820 int	 tcp_timer_active(struct tcpcb *, uint32_t);
821 void	 tcp_timer_stop(struct tcpcb *, uint32_t);
822 void	 tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int);
823 /*
824  * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
825  */
826 void	 tcp_hc_init(void);
827 #ifdef VIMAGE
828 void	 tcp_hc_destroy(void);
829 #endif
830 void	 tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
831 u_long	 tcp_hc_getmtu(struct in_conninfo *);
832 void	 tcp_hc_updatemtu(struct in_conninfo *, u_long);
833 void	 tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
834 
835 extern	struct pr_usrreqs tcp_usrreqs;
836 tcp_seq tcp_new_isn(struct tcpcb *);
837 
838 int	 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
839 void	 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend);
840 void	 tcp_clean_sackreport(struct tcpcb *tp);
841 void	 tcp_sack_adjust(struct tcpcb *tp);
842 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
843 void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
844 void	 tcp_free_sackholes(struct tcpcb *tp);
845 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
846 u_long	 tcp_seq_subtract(u_long, u_long );
847 int	 tcp_compute_pipe(struct tcpcb *);
848 
849 static inline void
850 tcp_fields_to_host(struct tcphdr *th)
851 {
852 
853 	th->th_seq = ntohl(th->th_seq);
854 	th->th_ack = ntohl(th->th_ack);
855 	th->th_win = ntohs(th->th_win);
856 	th->th_urp = ntohs(th->th_urp);
857 }
858 
859 #ifdef TCP_SIGNATURE
860 static inline void
861 tcp_fields_to_net(struct tcphdr *th)
862 {
863 
864 	th->th_seq = htonl(th->th_seq);
865 	th->th_ack = htonl(th->th_ack);
866 	th->th_win = htons(th->th_win);
867 	th->th_urp = htons(th->th_urp);
868 }
869 #endif
870 #endif /* _KERNEL */
871 
872 #endif /* _NETINET_TCP_VAR_H_ */
873